Mathematical Entertainments PDF

M. H.
Greenblatt
Mathematical Entertainments
A Collection of Illuminating Puzzles,
New and Old
Thomas Y. Crowell Company, New York
Established 1834
Copyright
1965 by M. H. Greenblatt
All rights reserved. No part of this book may

be reproduced in any form, except by a reviewer,
without the permission of the publisher.
Designed by Laurel Wagner

Manufactured in the United States of America
Library of Congress Catalog Card No. 65-15225

First Printing
To my wife, Nora
Introduction
My reasons for writing this book may very well contain a clue for its understanding and enjoyment.
First, there are a great many beautiful and elegant
facts and proofs in mathematics, which are published in
widely varying places, if at all. I would like to shout
about these problems from the housetops, but have chosen
the more dignified, and probably more efficacious, method
of writing a book about them.
Second, many people consider mathematics to be a
very boring, complicated subject. I hope to show that
some facets of mathematics can be quite simple and interesting, although they are not obviously so at first
glance.
Third, some of the most beautiful solutions of problems, of which I, alas, am not the originator, do not
appear anywhere to the best of my knowledge. Many I am
familiar with due only to the tenacious memory and
varied experience of M. W. Green.*
All of the problems and notes in this book have, at
one time or another, interested me very much, and I
would like to think the reader also may be fascinated by
them.
Some people may find that many problems appear to
be missing. Undoubtedly the collection is incomplete, but
all the problems I know of, and consider clever in some
* Research Engineer, formerly at R.C.A. Labs., Princeton, N. J.

vii
viii
Introduction
sense or another, are represented. Most of them can be

done mentally; none of them requires more than a single
sheet of paper. Some of the items treated here are well
known, but presented in a different way to illuminate a
different facet. I earnestly hope this book will be received
as it was intended, namely, to furnish pleasant relaxation
and to show that mathematics can be beautiful.
No preface is complete without thanks being expressed to the "person without whom this book could not
have been written." In the present case, I wish to thank
my wife.
M. H. GREENBLATT
Contents
I: "Census-Taker" Problems
The Neighborhood Census, The Priest and the

Banker, The Three Martians
2: Diophantine-Type Problems
The Missionary and the 159-Link Chain, Restored Arithmetic Problems, Block Cut into
Cubes, Basketball Scores, The Incorrectly
Cashed Check, On Moving Day, Expression
for Which 2,592 Is a Solution, Railroad Trestle
Problem, Integral Pythagorean Triplets
3: Konigsberg Bridges, Design Tracing, and Euler 18
Story of Seven Bridges, Design Tracing, Parity
of Number of People Shaking Hands at a Conference, Euler's Formula
4: Bits and Pieces
25
Volume of an Anchor Ring, Topological Behavior of a Loop Around a Torus, Which Raise
Nets More Money? Stacking of Discs, Pecuix
Contents
liarity of 142,857, Hempel's Paradox, Geo-
Nm 3
metric Proof That ~
[N~ m J2 , A Peculiar
Infinite Sequence, Wilkes-Gordon Construction, Proof of the Pythagorean Theorem, The

Advantages of Compound Interest
5: Coin Weighing Problems
43
Determination of Odd Coin, Weak and Strong

Solutions, A Strong Solution, Another Coin
Problem
6: Fermat's Theorem; Sailors and Coconuts
49
Three Men and Twelve Lumps of Sugar,

Sailors and Coconuts, a5 - a Is Divisible by 30,
Divisibility of a 13 - a, Proof of Fermat's
Theorem
7: Partitioning Problems
61
The Maid at the Stream, Division of Balsam

Among Three People, Another Urn Problem,
Pile of Bricks, Poker Chips, Finding Any Rational Fraction of a Volume V, The Euclidean
Algorithm
8: Symmetry
Sam Loyd's Daisy Puzzle, Cigars on Table,

Cup of Coffee and Cup of Tea, Nim, Nim",
The 21 Counters, Grundy's Game, Mutilat<>d
Chess Board, Construction of a Cube
71
Contents
9: Parlor Tricks and Number Manipulation
xi
82
Eight 8's, The Multiplication Theorem, Some

Useful Facts, Casting Out Nines
10: Divisibility by Various Numbers
92
Divisibility by 2, 3, 5, 7, 9, 10, 11
11: Problems Which Seem to Have No Solution
99
The Careless Indian Who Drops His Paddle,

The Early Commuter, The Officer and His
Column of Men, The Master Painter vs. the
Apprentice, The Two Ferry Boats, Volume of
a Modified Bowling Ball
12: Probability Problems
106
Die Throwing, Bertrand's Box Paradox, Probability of Identical Pairs Among Cards, Probability of Same Birthday Among N People,
St. Petersburg Paradox
*. (Read It to Find Out)
114
14: Humor in Mathematics
115
Liars and Truth-Tellers, Liar or Truth-Teller

-at N-tuple Intersection, Where on Earth
Was He? Goldbach's Conjecture, A "Seemingly Impossible" Problem, Archimedes' Cattle
Problem, Definition of a Catalyst, Mnemonics
for the Value of Pi, A Short History of the
Accuracy of Pi, Best of All Possible Mnemonics, Experimental Determination of Pi, A
xii
Contents
Letter in the Physical Review, Phi: The

"Golden Ratio," Probability That Two Random Numbers Are Relatively Prime, and
Euler's Theorem
15: A Slight Electrical Interlude and Some Undigested Thoughts
128
Two-Dimensional Array of Resistors, Ten

Wires Under the River, Perimeter to Enclose
Maximum Area
16: Left-Overs-From Greeks to Computers
135
A Computer Correlation, Translation of Languages by Computer, Fair Division of Cake

Among N People, The Trolley Cars: An Apparent Violation of Probability, Napoleon's
Problem, The Mohr-Mascheroni Construction,
Random Chord of a Circle, Description of
Greek Names for Numbers, Some Geometricians Are Masochists, Ancient, Admirable Calculations, Fermat's Challenge, Dance of the
Bugs, The Swiss Mountain Moving Company,
Ltd.
Index
155
Mathematical Entertainments
1: "Census-Taker" Problems
One of the few amusing things to come out of World War

II was a new type of brain twister-the "census-taker"
problem. (The time and place of origin of a problem are
difficult to specify. To the best of the author's knowledge,
this problem was born on the M.LT. campus in one of
the war projects.) Several different versions have appeared,
three of which I have labeled the "Neighborhood Census,"
"The Priest and the Banker," and "The Three Martians."
The Neighborhood Census
A census taker, who was very intelligent, but eccentric in that he never asked more questions than necessary,
came to a house where three of the inhabitants were not
at home. To get their ages, he asked the housekeeper,
"What is the product of their ages?"
The housekeeper answered, "1,296."
"What is the sum of their ages?"
"It's the same as the house number." (At this point,
we must note that the census taker knew the house number, but you, the problem solver, do not.)
The census taker then said, "I still can't tell their
ages. Are any of them older than you?"
\Vhen the housekeeper answered, "No," the census
taker said, "Now I know all their ages!"
1
"Census-Taker" Problems
What are those ages?

At first glance this problem appears so "patently
incomplete" that a solution seems impossible without invoking necromancy or witchcraft. (My first acquaintance
with this problem came in a letter from a friend_ I was
convinced that his description was incomplete, and I
answered by tactfully suggesting a place where the little
men in white coats are kind and gentle.) And yet, a logically consistent derivation of a unique set of ages does exist.
In this problem, the product of the three numbers
and the sum of the same three numbers were given to
the census taker. This information is often sufficient to
nl
n2
na
Sum
nl
n2
na
1
1
1
1
1
1
1
1
1
1
1
1
1
2
1
2
3
4
6
8
9
12
16
18
24
27
36
1,296
3
4
6
8
9
12
18
2
3
3
3
3
3
3
3
3
4
4
4
4
4
6
6
6
6
8
9
9
24
3
4
6
8
9
12
16
18
4
6
9
12
18
6
8
9
12
9
9
12
27
144
108
72
54
48
36
27
24
81
54
36
2
2
2
2
2
1,298
651
436
329
223
171
154
121
98
91
79
76
73
328
221
168
116
91
83
68
56
2
2
648
432
324
216
162
144
108
81
72
54
48
36
324
216
162
108
81
72
54
36
Table 1
27
18
36
27
24
18
18
16
12
Sum
53
150
115
81
65
60
51
46
45
89
64
49
43
40
48
41
39
36
35
34
33
"CensusTaker" Problems
determine the three numbers. In this case, the census

taker confessed that he still was unable to tell the three
ages. Therefore, of the groups of three numbers having
the product 1,296 there must be at least two groups with
the same sum. All groups of three factors of 1,296 are
shown in table 1, along with the sums of the groups.
Except in one case, the sum of each group of three
numbers is different from the sum of all other groups.
This exception is 91, the sum of 1, 18, and 72, and also
of 2,8, and 81. Had the house number been any allowable
sum other than 91, the census taker would have been able
to choose the unique triad for the ages of the three missing
people. But he couldn't. So the house number must have
been 91.
Since the last question enabled him to decide which
of the two triads was the correct one, the housekeeper
must have been between 72 and 81. (Another fact known
to the census taker but not to you. For only then would
the answer "No" enable the census taker to announce that
he knew all the ages-namely 1, 18, and 72.)
The Priest and the Banker
The town priest saw his friend the village banker

walking down the street with three girls. Not known for
excess verbiage, the priest asked, "What is the product
of the ages of these girls?"
The banker answered, "2,450."
"The same as yours, Father."
"I still can't tell their ages. Are any of them as old
as you?"
When the banker said, "No," the priest announced
that he knew all the girls' ages.
What are those ages, how old is the priest, and how
old is the banker?
The solution to this problem is very similar to that

for the "Neighborhood Census," but there are a few new
wrinkles. The product of the girls' ages is 2,450. The factors
of this number are 1, 2, 5, 5, 7, and 7. Again it is helpful
to construct a table showing all the triads that have the
product 2,450. Table 2 shows all such triads, along with
their sums:
nl
n2
na
Sum
nl
na
Sum
1
1
1
1
1
1
1
1
1
2
1
2
5
7
10
14
25
35
49
5
2,450
1,225
490
350
245
175
98
70
50
245
2,452
1,228
496
358
256
190
124
106
100
252
2
2
2
5
5
5
7
7
7
7
25
35
5
7
10
7
10
14
175
49
35
98
70
49
50
35
25
184
76
72
108
82
64
64
52
46
Table 2
We can see that only 5, 10, 49 and 7, 7, 50 have the

same sum, 64. The priest's age must be 64. Only if the
banker was 50 would the priest have been able to choose
the correct triad. Therefore, the banker is 50, and the
girls are 5, 10, and 49.
The Three Martians
The last twist of this popular type of problem is, in

a sense, the trickiest, but still logically consistent. It is
said to have been brought back from Mars by an Air Force
colonel who visited there. According to him, the Martians
are a very intelligent people, partly because they have
two heads, as shown in this reproduction of a sketch he
brought back.
Figure 1
Wanting to know the ages of three Martians, the

colonel asked the Chief of Martian Intelligence, "What
is the product of their ages?"
"1,252" was the answer.
"The same as my father's."
"I still can't tell their ages. Are any of them as old
as you?"
When the chief said, "No," the colonel announced
that he could tell the ages. What are they?
This problem seems so similar to the previous ones
that the natural urge is to plunge right in and prepare a
table of triads. The factors of 1,252 are 1, 2, 2, and 313.
nl
n3
Sum
1
1
1
2
1
2
1,252
626
313
313
1,254
629
318
317
Table 3
When we inspect this table, we are tempted to exclaim, "Mon Dieu, has Greenblatt flipped or lost his
marbles? No two triads have the same sum!"
"Census-Taker" Problems
But, after our passions have subsided, we decide to

have another look at the problem. We know we are using
the right method to find the answer. We think-perhaps
there is something wrong with the numbers! We note from
the sketch that the Martian has three fingers on each
hand. And, with logic worthy of the great Sherlock Holmes
himself, we deduce that the Martians must use the number
base 6 because they have six fingers, just as we use the
base 10 because we have 10 fingers (5 fingers on each hand).
The number 1,252 in the base 6 is equal to:
1(63) + 2(6 2) + 5(61) + 2(6)
= 320 in the base 10 (also written 32010)
It is immaterial whether we form the table of triads
in the base 6 or the base 10, because triads having the
same sum in one base will also have the same sum in the
other base. We construct table 4 in the base 10 because
most of us are more familiar with it. The factors of
32010 are 1, 2, 2, 2, 2, 2, 2, and 5.
nl
n2
na
Sum
nl
n2
na
Sum
1
1
1
1
1
1
1
2
1
2
4
5
8
10
16
2
320
160
80
64
40
32
20
80
322
163
85
70
49
43
37
84
2
2
2
2
4
4
4
5
4
5
8
10
4
5
8
8
40
32
20
16
20
16
10
8
46
39
30
28
28
25
22
21
Table 4
The fact that the colonel had to ask one more question
indicates that the three ages must have the same sum.
The triads 2, 10, 16, and 4, 4, 20 both have the same sum,
28. Since the colonel was able to choose between the triads
because the chief said none of them was older than he,
the triad in question must have been 2, 10, 16. In the

number base 6, the decimal triad 2, 10, and 16 would be
26, 148 , and 246, and the decimal triad 4, 4, and 20 would
be 46, 46, and 326. The chief's age had to be between 16
and 20 Martian years (or between 246 and 326 written in
his own number system). A Martian year, incidentally, is
nearly twice as long as an Earth year, and this accounts
for the fact that the chief was so young-less than 32
(Martian years).
2: Diophantine-Type Problems
Diophantine equations are those equations in which only

solutions in integers are permissible. For example, the
equation x 2 + y2 = Z3 is a Diophantine equation when
each of the variables is an integer. One solution is:
x = 2,
= 11, and z = 5
22
+ 112 = 53
which yields:
There is another solution to this equation in which x and

yare the same, namely:
x = 2,
= 2,
= 2
which leads to:

22
+ 22
23
These are the only solutions (excluding the trivial x = 0,

y = 0, Z = 0).
Diophantine equations are vastly different from equations which are not restricted to integral values of x, y,
and z. The original equation, x 2 + y2 = Z3, would have an
infinite number of solutions in continuous variables, but
only two with non-zero integral values.
Many of the problems in the following chapters obviously require integral solutions. In this chapter, we will
consider other problems which are soluble only by integers.
8
Diophantine.Type Problems
The Missionary and the 159Link Chain
One of the simplest problems, which requires no mathematical analysis at all, is about a missionary who had a
gold chain with 159 links. He was captured by a group of
cannibals (as are most missionaries in mathematical problems), who stipulated that he could purchase one day of
freedom for each single additional link of the gold chain
that he gave them. They agreed to accept the principle
of "change-making" (i.e., he might give them a 7-link
section, if they had 6 links to give him in change). The
missionary's goldsmith charged an exorbitant rate for making each cut link whole again. What is the minimum
number of links that he had to cut in order to assure his
existence for 159 days (at which time he knew that he
would be rescued)?
The solution to this problem is simplified if we recognize that by cutting a single link, he could obtain two
smaller chains and a single cut link. By making only four
cuts, he produced small chains of length 5, 10, 20, 40,
and 80, and four single links. These enabled him to live
until the plane came.
Restored Arithmetic Problems
The previous problem hardly requires Diophantine

analysis. A better example of an elementary Diophantine
problem might be one in the class known as "restored
arithmetic." In these problems some or all of the digits
have been erased and are represented by dots or letters.
The problem solver is asked to find digits consistent with
the rest of the problem. The answer requires the solution
of many small Diophantine equations. Some information
is always given in these problems, and it is natural to
wonder how" degraded" this information can be. Problems
in which only one digit is given are known. W. W. R. Ball
10
gives an example* of a restored division problem with no

digits given. The only information is that the quotient of
the problem is a repeating decimaL
We can make up a simple problem: ~. Please note
that this seems to contain no information at all, until we
state that it is a fifth root extraction. The only two-digit
integer whose fifth root is a single integer is 32.
Figure 2 shows an original problem in restored square
root extraction. The only information given is that the
answer is an even number. The answer is 8,604.
Yxx
xx xx xx
xx
xx xx
x xx
x xx xx
x xx xx
Figure 2
Rectangular Block Cut into Unit Cubes
The next problem is: A rectangular parallelepiped,

of dimensions abc, where a, b, and c are integers, is painted
red on the outside. The block is then sawed into unit cubes,
and it is found that the number of cubes having no paint
at all is exactly equal to the number of cubes having some
paint on them. What are the allowable values of a, b, and c?
The total number of unit cubes is abc, and the number
of cubes in the" inner core" (which have no paint on them)
is the volume of the rectangular parallelepiped one unit in
* Ball, W. W. R., Mathematical Recreations and Essays, New

York: The Macmillan Co., 1962, 11th edition, pp. 22 ff.
11
from each edge, or two units less in each dimension. We

therefore want solutions for which
abc
= 2[(a - 2)(b - 2)(c - 2)]
An 8 X 10 X 12 original block will have an inner core

of size 6 X 8 X 10. The ratio of these two volumes is
accordingly 8 X 10 X 12 divided by 6 X 8 X 10. The
eights and the tens cancel each other, and we are left with
a simple ratio: which, indeed, equals 2.
Further solutions cannot be seen in this simple manner.
A straightforward way to derive these other solutions L~
to set up the equation:
abc
= 2
(a - 2)(b - 2)(c - 2)
Dividing by abc, and inverting, we obtain:
Now we can try various integers for a, b, and c. Suppose, for example, we arbitrarily choose a = 5. Then the
first fraction in parentheses is equal to 1. To get rid of
the 5 (so that we may ultimately obtain the fraction !)
we must choose for b a number which is 2 greater than a
multiple of 5. Twelve is one such number, but it is not
large enough, because the second parenthesis would be
H- (t) The first two fractions are now equal to 1 and t,
and the product is already equal to !, so that pair is not
allowable. Let us try b = 17. The fraction in the second
parenthesis is now it, and the total fraction so far is 197 '
still more than !. We must now try to multiply this fraction by a suitable (c - 2)/c so that the product will equal
!. Again, we see that c = 36 will have a very happy result
because (c - 2)/c = H and this fraction, multiplied by
ft, equals t or !. This method of solution is the best f hat
I can offer. These Diophantine problems do not haw the
12
Diophantine-Type Problems
same quality of generality that I try to stress in the succeeding chapters. I feel almost apologetic about including
them, but they are interesting. Perhaps some reader will
be able to furnish the generality that my exposition lacks.
Some additional values for a, b, and c are given in
table 5. There are about ten more solutions, and the reader
a
5
5
6
6
6
6
7
8
8
8
13
18
132
32
32
24
20
16
100
18
14
12
10
11
12
14
7
8
9
10
Table 5
may amuse himself by finding them if he wishes to. The

underlined values are thought to be the only sets consisting
solely of even numbers.
Basketball Scores
A basketball team thought one point for a free throw

and two points for a basket weren't really commensurate
with the difficulty in achieving them. So they decided to
award a points for a free throw, and b points for a basket.
After this was done, they discovered there were precisely
35 scores that could not be attained, and one of these
scores was 58. What were the values of a and b?
We will not even consider the pedestrian method
of simply trying different numbers for a and b. After
DiophantineType Problems
13
playing around with this problem and several sets of

numbers, we discover that a and b must be relatively
prime (have no common factor). We can also discover
that (a - l)(b - 1)/2 scores cannot be achieved.
The proof of this statement will not be given but is
"left as an exercise for the interested reader." With the
information given we can write that (a - l)(b - 1)/2 =
35. Knowing that 58 is one of the unattainable scores,
deep concentration and soul searching will show that 8
and 11 are a suitable pair of numbers. Further investigation indicates that these form a unique solution. In this
example, as in the last, the lack of generality in the solution of Diophantine equations is quite apparent.
The Incorrectly Cashed Check
Our next problem is about the man who received a

paycheck for so many dollars and so many cents. He
cashed the check at a bank where the teller made the
unlikely mistake of giving him as many dollars as there
should have been cents and as many cents as there should
have been dollars. Our hero did not notice the error, but
went forth and made a five cent telephone call (you can
see how old this problem is). He then discovered that he
had precisely twice as much money left as the check had
been written for. What was the amount of his paycheck?
This problem allows a somewhat more rigorous solution than the previous ones. We recognize, first of all, that
the number of dollars on the original check must have
been less than 100 for the teller to have made the mistake
which we say he did. If we assume that the original check
was for r dollars and s cents, the incorrect sum must have
been (l00s + r) cents. Now we can write the equation:
100s + r - 5 = 2(100r + s)
lOOs+r-5=200r+2s
199r - 98s + 5 = 0
14
must be approximately twice as large as r, so let us

write 8 = 2r + x where x is some, hopefully small, correction. Substituting this new value for 8, we have:
+ +5 = 0
+5 = 0
+5 = 0
199r - 98(2r
x)
199r - 196r - 98x
3r - 98x
3r=98x-5
We know that r is a whole number less than 50. Therefore, the expression 98x - 5 (which is equal to 3r) must
be less than 150 and must be divisible evenly by 3.
If x were equal to 2, 98x - 5 would be greater than
150. If x were equal to 0, 98x - 5 would be negative-an
impossible solution. Therefore, x must be equal to 1.
Continuing:
3r = (98 X 1) - 5, or 93
r = 31, 8 then equals 63
The original check must have been for $31.63
On Moving Day
A friend of mine, while we were graduate students,
was quite a decent student, but was extremely absentminded. However, he wasn't quite as bad as the hero of
the following anecdote. A late, very well known physicist
was exceptionally brilliant in technical matters, but was
helpless when it came to matters concerning the running
of a house. So bad was this situation that his wife had to
run all household affairs. One day, they were going to
move into a new house. When he left for work that morning, his wife asked him to please remember to come home
to their new house, since no one would be home at the
old house. In order to have a story, he returned to the
old house, of course. He strode up to the front door, firmly
expecting to find it unlocked. When it refused to yield,
he was very puzzled. He went around to the side and
15
peered in a window. Seeing the house completely empty,

he said, "What a fool I've been! My wife told me this
morning that we were moving to the new house, and I
should have remembered. But now-I wonder where we
moved to?"
Just then he saw a small girl walking down the street.
With a flash of inspiration, he stepped to the sidewalk
and said to her, "Pardon me, little girl. The Browns used
to live in this house, but I understand they moved today.
Can you tell me where they moved to?"
"Yes, Daddy," she replied. "Mommy sent me here
to get you!"
Expression for Which 2,592 Is a Solution
This friend of mine from graduate school was nearly

as absent-minded as Dr. Brown. One day he came into
our room while Larry and I were minding our own business
(studying, of course) and asked, "Say, fellows, how am I
going to remember my home phone number? It's EVergreen 2592.* I can remember the exchange all right, but
the numbers confuse me. Isn't there some easy way for
me to remember them?"
I considered the question an idle one and promptly
forgot about it. Several hours later, Larry, drawn, haggard,
and with bloodshot eyes, started muttering to himself,
"No, it can't be, it isn't, but it is!" Then he gleefully
explained, "The number 2,592 appears to be a unique
solution in integers of the expression x,yzx = x"z"', where
the x, y, and z represent the same digits in the same order
on both sides of the equation (a decimal expression on the
left and the powers of integers on the right), and are
different from each other."
* This is a true story. The phone number in question was the

phone number of B. Rothlein, who lived on 41st Street in Philadelphia between 1943 and 1947.
16
Larry's method for discovering that 2,592 is a unique

solution of the above expression is completely unknown
to me. I'm not even sure of a concise way to prove its
uniqueness.
Railroad Trestle Problem
Two boys were crossing a railroad trestle when they

suddenly heard the sound of an approaching train. Being
prudent, they started to run for safety toward opposite
ends of the bridge. Running at the same speed, each
reached his particular end of the bridge just in time to
avoid being hit by the locomotive. The boys werc i of the
way across when they heard the sound of the whistle. The
train was going 50 mph. How fast did the boys run?
In solving this problem, we must recognize that the
time taken by the train to cross the bridge was equal to
the difference in times at which the boys arrived at their
respective ends. If the boys were i of the way across when
they heard the whistle, then one had a distance iL to run
(where L is the length of the bridge) and the other had to
run !L. The difference in the distances that the boys
ran is tL, and if they ran at a velocity VB, then the time
difference was tLjV B = Lj5VB. This must be equal to
the time that the train took to cross the bridge, or LjV T.
Equating these two expressions:
5V B
VT
-- = -
or
The boys therefore ran at a velocity of 10 mph.

Integral Pythagorean Triplets
The discovery of integral Pythagorean triplets can be

fascinating. (Pythagorean triplets are integers a, b, and
c, which are related by the expression a2 b2 = c2 If the
Diophantine-Type Problems
17
value of c is given, there are infinitely many solutions for

a and b in nonintegers, but there may be only a few in
integers.)
The squares of two consecutive integers, nand n + 1,
differ by 2n + 1. If 2n + 1 is, itself, a square, we will
have the Pythagorean triplet V~n + 1, n, n + 1. For
example: if 2n + 1 = 25, then we would have the Pythagorean triplet 5, 12, 13 because the difference between 122
and 13 2 is 25. An infinite set of Pythagorean triplets can
be built up in this way.
An interesting problem can be mentioned here. I once
casually asked my children how many integral Pythagorean triangles they could name, in which one edge was
15 units long. The answer is that the 3, 4, 5 right triangle
can be multiplied by 3 to give a 9, 12, 15 right triangle,
and by 5 to give a 15, 20, 25 right triangle. Then, since
15 2 is 225, we can write: 225 = 112 + 113, and so 15, 112,
and 113 is such a triangle. By somewhat more exotic
methods, we discover that 8, 15, 17 is such a triplet. We
can multiply the 5, 12, 13 triangle by 3 to get 15, 36, 39.
The answer is that there are five such triplets.
If we choose any two numbers, u and v, then the
number u 2 - v 2 and the number 2uv form the two lower
numbers of a Pythagorean triplet. The other number is
u 2 + v2 The reason is that (u 2 - V 2)2 = u 4 - 2U 2V 2 + vt,
(2UV)2 = 4u2v 2, and the sum of these two is:
u 4
+ 2U2V 2 + v 4 =
(u2
+ V 2)2
All primitive Pythagorean triplets can be formed this way.

A more difficult problem has been suggested in the
American Mathematical Monthly (1914). It asks to find
three integers such that the sum of the squares of any two
of them is, itself, a perfect square. The solution to this
problem is involved, but not difficult. The smallest numbers are 44, 240, 117, and we refer people interested in a
demonstration to the original source.
3: Konigsberg Bridges, Design Tracing,

and Euler
Story of Seven Bridges
In Germany, there is a town called Konigsberg, situated on the banks of the Pregel River. The island of
Kniephof is in the river, just below another island, and
both islands and the banks are connected by bridges as
shown in figure 3:
PREGEL RIVER
These are the seven Konigsberg Bridges. And thereon

hangs a tale. The residents of Konigsberg used to wonder
if it was possible to walk and cross each of these bridges
once and only once. Try as they would, they never succeeded, but they didn't know if their failure was due to
their limited imaginations or to the nature of the problem.
Word of their dilemma finally reached the ears of the
great mathematician Leonhard Euler. He offered a solution to the problem which is beautiful in its simplicity.
18
Konigsberg Bridges, Design Tracing, and Euler
19
First he replaced the islands and land masses by points,

and the bridges by lines-as in figure 4.
A
Figure 4
.--------------..;;;.0
Following Euler's method we can characterize each

junction of lines (or bridges) as odd or even, depending
on how many lines emanate from it. If all junctions are
even, we can traverse the entire array, going over each
line once only, because for every line entering a junction,
there will be another line leading out. If there are, at
most, two odd junctions, the complete traversal can still
be made, provided we start at one of the odd junctions
and end at the other. If the array has more than two odd
junctions, then no power on earth can accomplish the
traversal in one continuous path. In figure 4, all four
junctions are odd. (The number of odd junctions, incidentally, must always be even. The proof of this is very
similar to the proof that the number of people shaking
hands at a conference an odd number of times must be
even. This problem is considered later in this chapter.)
If we insert another bridge from A to B, two of the
junctions (A and B) will become even, and the other two
will remain odd. A complete journey could then be made
by starting at Cor D. Without the additional path from
A to B there are four odd junctions, and therefore all the
20
paths can not be traversed in a single journey. All paths

can, however, be traversed in n/2 separate journeys, where
n is the number of odd junctions. The Konigsberg Bridges,
therefore, can not be traversed in a single journey, but they
can be traversed in two separate journeys.
Design Tracing
The solution of the Konigsberg Bridge problem is

directly applicable to a class of problems known as design
(a)
(e)
(b)
Figure 5
(d)
21
tracing problems. Figure 5 shows some typical examples

of drawings for this type of problem.
In each case, the question is: Can this design be traced
in one continuous line (i.e., without lifting the pencil
from the paper, or retracing any line)? Problems such as
these can always be done if there are, at most, two odd
vertices. Otherwise, it would take n/2 distinct lines or
tracings. The drawing in figure 5(a) has eight odd vertices,
(b) has two odd vertices, (c) has two odd vertices (the
ends of the vertical line), and (d) has twelve odd vertices.
Part (a) requires that we lift the pencil four times. Part (b)
can be drawn in one continuous line if we start at one of
the lower comers and end at the other lower corner. Part
(c) can also be drawn in one line if we start at the top or
bottom. Part (d) requires six separate tracings.
Slight variations on this design tracing problem are
popular. They are usually subject to the same analysis
as we have considered. For example, a problem often
given is shown in figure 6:
.r. I
I ~~
..
Figure 6
~~II I~~
..
The outline shows a five-room diagram. There are

doors as shown in the drawing. The question is: Can a
walk be planned so as to pass through each doorway once
and only once? Inspection shows that rooms A and C
have an even number of doors leading into them, but
rooms B, D, and E have an odd number of doors leading to
them. So the required walk cannot be performed. If the
top wall in room B had another door, then there would
be only two odd rooms, and the walk could be performed,
if we start in D or E (and end in E or D).
22
Weird stories concerning" a person that a friend of

mine knows," who has solved a problem such as the one
in figure 6, occur very frequently. Invariably, the time
of the day, or the season, or the phase of the moon, or
unsettled world conditions are such that we cannot see
the "solution." The existence of such stories must be
accepted as par for the course in this field of impossible
constructions.
Whenever an outline can be traced in one continuous
line, that line can be drawn so that it never crosses itself.
A method for doing this has been proposed by T. H.
O'Beirne. A full explanation can be found in the April
1964 issue of Scientific American. *
Parity of Number of People Shaking Hands
at a Conference
Another problem susceptible to the treatment given

to the preceding problems is the one about people shaking
hands. At a recent conference, the members shook hands
with each other an arbitrary number of times. Prove that
the number of people who shook hands an odd number
of times must be even.
The answer is easy to see if we consider a drawing in
which each person is represented by a dot, and each handshake by a line joining two dots. Each dot can be labeled
o or E, depending on whether there are an odd or even
number of lines coming from it. Each new line that is
drawn must necessarily go from an 0 to an 0, or an E to
an E, or an 0 to an E. A line going from 0 to 0 changes
both of them to an E and thus decreases the number of
O's by an even number-2. A line going from E to E
changes both to an 0, and this, again, changes the number
of O's by 2. A line from E to 0 changes the E to an 0,
and vice versa-the number of O's is unchanged. Since
the number of O's can only change by an even number,
* M.
Gardner, Scientific American, April 1964, p. 126.
23
and the process must start with zero O's (an even number)
-the number of O's must remain even.
Euler's Formula
Another beautiful derivation by Euler relates the number of edges (k,), the number of countries (C), and the
number of vertices (V) in any map on a plane surface.
(The formula relating E, C, and V on the surface of a
sphere is almost identical.) This formula can be presented
in a very simple manner-a way which makes hardly any
mathematical demands beyond a keen interest. Consider
the map in figure 7a:
Figure 7a
Let us consider the area outside the group of countries, labeled g, to be an ocean. Each of the boundaries is
a dike, which prevents the country from being flooded.
Now break down the dikes systematically, so as to flood
each country, one at a time. But do not break any dikes
except to flood a new country. The unbroken dikes are
then all connected to each other. (For if two groups were
not so connected, then the last dike joining the two groups
should not have been broken, because there was water
on both sides, and no new country was flooded thereby.)
Figure 7b shows the same map as figure 7a, but dotted
lines indicate the dikes broken to flood each country. In
addition, a large dot is placed at each vertex where three
or more edges met.
24
,-- .
,
' - -
e _ - - - -____________________
Figure 7b
In our derivation, we will represent the number of

broken dikes by B, and the number of unbroken dikes by
U. The number of broken dikes must exactly equal the
number of countries, C. For at least one dike per country
must be broken, and there is no need to break more. The
unbroken dikes are all connected, as was mentioned. And
if we start at a vertex such as the one labeled 1 in figure 7b,
then each unbroken dike will have an additional vertex
at its end. Thus, we have B = C, and V = U + 1. Since
the total number of edges (or dikes) must be equal to
B + U, we can write:
E=B+U=C+V-l
This is Euler's formula. On the surface of a sphere,
the same relation holds, except that C increases by one,
since g in figure 7a is another country. To account for
this, the formula changes to E = C + V - 2. This formula
forms the basis for a proof that there are no more than
five regular polyhedra (the so-called Platonic solids). It
is also used in a proof that five colors are sufficient (but
not that they are necessary) to color any map, so that
each country may be colored differently from countries
adjacent to it. It can still not be proved that four colors
are sufficient on a plane. (On the surface of a torus, or
doughnut, it can be proved that seven colors are both
necessary and sufficient.)
4: Bits and Pieces
Volume of an Anchor Ring
How can we determine the volume of an anchor ring,

as shown in figure 8?
Figure 8
I+I--A
Pappus' Theorem, which is a part of integral calculus,

can be used to determine the volume. But the method
described here is considerably simpler, and exact. Consider the ring cut in half as shown in figure 9(a), and
the lower half rotated from the position indicated by the
dotted lines.
Then, rotate the portions outside the vertical lines
shown in figure 9(a), so that the ring now resembles
25
26
Bits and Pieces
Figure 9
figure 9(b). Again, the portions at each of the lines in

figure 9(b) can be rotated to resemble figure 9(c). At
each step, the volume of the original anchor ring is not
changed at all-it is merely rearranged. The process described will eventually result in a straight circular cylinder
whose cross-sectional area is 1r1"2, and whose length is the
circumference of the dotted line at A in figure 8, namely
211'"(a + b)/2. The ends of the cylinder are perpendicular
to the axis of the cylinder because they were originally
cut "square" and were never changed after that. The
volume of the resultant tube is therefore:
Q.E.D.
Bits and Pieces
27
Topological Behavior of a Loop

Around a Torus
The next problem could belong in a chapter in which

the obvious solutions are wrong. Consider a rubber inner
tube. Can this tube be turned inside out through a hole
at the outer edge, and still be, topologically, an inner
tube? The simple, but treacherously incorrect, solution
considers an inner tube as shown in figure 10. A string is
Figure 10
wrapped around the smaller diameter as is shown at A.

It is inextricably looped with the tube, and no simple
deformation can change its "loopedness." The only way
to unlink the string from the tube would be to cut one of
them. But if it were possible to turn the tube inside out
and still have a tube, then the string at A would be on the
inside, and one would assume that we could reach in
through the hole and pull out the string. Since the" loopedness" cannot be so easily changed, it seems impossible to
turn the tube inside out and still have a tube. But as
stated earlier, this argument is fallacious.
Figure 11 shows the inner tube in six stages during
the deformation.
Figure Il(a) shows the inner tube before deformation.
In (b), the hole shown at the left of the drawing in (a) is
tremendously enlarged. In (c) and (d) the edges of that
28
Bits and Pieces
(A)
(0)
( C)
(F)
(E)
Figure 11
Bits and Pieces
29
hole are folded back, and if the large circle is now

allowed to shrink back to its original size, as in (e) and
(f), the operation is complete. (f) resembles (a) if (a) is
rotated 90 about a horizontal axis. But the string which
was tied around the" small" diameter is now around the
"large" diameter-around the inside of the tube! This is
why the original argument was wrong. It is possible to
turn an inner tube inside out and still, topologically, have
an inner tube.
Which Raise Nets More Money?
The next problem can be relied upon to start an argument. A man is offered a job at $1,000 a year, and is
offered increases of $50 more each half year or $200 more
each year. Which choice will net more money? The choice
seems so obvious that there must be a "hooker" somewhere. But, I assure you
(a) there is no hooker, and
(b) the obvious solution is wrong.
Let us construct a table showing the amounts of
money earned in each case:
$50 raise
each half year
1st half year
2nd half year
Total, 1st year
500
550
3rd half year

4th half year
Total, 2nd year
600
650
5th half year

6th half year
Total, 3rd year
700
750
Total received over 3 years
Table 6
$00 raise
each year
1,050
1,000
1,250
1,200
1,450
1,400
$3,750
$3,600
30
Bits and Pieces
The raise of $50 every half year is actually more

rewarding than the alternative. (I don't think that this
problem should be used as a proof that" all that glitters
is not gold.")
Our intuition may have led us astray because we
didn't take time to analyze the problem. Everybody ean
figure out what "$50 more every half year" means in
terms of a year, but few people recognize immediately
that $50 per half year is the same as $100 for a full year,
and further a raise at this rate occurs twice a year! Thus,
the two alternatives are raises at the same yearly rate,
but the half-yearly raise starts earlier.
Stacking of Discs
You are given a large number of discs, and asked to

place them one on top of the other, displacing each one
horizontally by as much or as little as you please. What
is the maximum horizontal displacement of the top disc
from the bottom disc you can obtain before the stack
topples?
This is another problem for which an automatic reaction can mislead. The usual answer is that the top disc
can surely not be displaced by more than one radius (or
maybe, a diameter) from the center of the bottom disc.
As is customary in this book, the instinctive answer to a
simple question is wrong.
The simplest way to visualize the correct solution is
to consider the arrangement of the discs from the top
going down. The drawing in figure 12 shows the arrangement of the top six discs, each being horizontally displaced
as much as possible.
The radius of each disc is r, and the center of gravity
of the disc is represented by the dot in the center. The
edge of disc 2 is placed under the dot of disc 1. The center
of gravity of the first two discs is now r/2 in from the
Bits and Pieces
31
Figure 12
edge of disc 2 (as is shown by the arrow on the lower

edge of 2) and the edge of disc 3 must be put under that
point. The center of gravity of any particular group of
discs must, for stability's sake, be placed on the next disc
below it. In order to achieve the greatest possible displacement, it is most advantageous to put it directly at the
edge. The center of gravity of the three discs is r/3 in
from the edge of disc 3, and the edge of disc 4 must be
placed under that point. The edge of disc 5 is displaced
by r/4, and that of disc 6 by r/5, etc. The displacements
form the series r, r/2, r/3, r/4, etc., and the total displacement (where n is the total number of discs) is thus:
Displacement
r(l
+ ! + I + t + ... + l/n)
This is the harmonic series, which is known to diverge.

The top disc can thus be displaced from the bottom one
by as much as we wish!
It is amusing to assemble ten pennies in this manner.
The sum of the first ten terms of the harmonic series is
nearly 3, and with only ten pennies, the bottom penny
can be completely out from under the top one.
Peculiarity of 142,857
Consider the number 142,857. If it is multiplied by

2 or 3 or 4 or 5 or 6, the result is a cyclic permutation of
the original number. For example, 142,857 X 3 = 428,571.
The usual question is: Can this be easily explained? It
can, and the reason is that 142,857 is the repeating portion
32
Bits and Pieces
of the decimal expansion of the fraction t, and the periodicity (6) is one less than the reciprocal of the fraction
itself (7). This means that the remainders, on performing
the required division, must have six different values. (If
the remainders were ever the same as a previous one, then
the repeating portion of the decimal would start right
there.) But since the repeating portion of the decimal has
a periodicity of 6, and each remainder must be less than
7 and cannot be 0, each number from 1 to 6 must occur
just once per cycle in the sequence of remainders. When
the remainder is 2, then the sequence of numbers in the
quotient is exactly the same as it would be if we were
expanding t. And similarly, when the remainder is 3, the
sequence must be the same as that for --. So the sequence
for an integral number of sevenths must be the same in
each case.
Seven happens to be the only single digit (not counting
2) whose reciprocal is a repeating decimal of periodicity
one less than the number. There are other numbers whose
reciprocal has a periodicity one less than the number itself,
but these numbers are large, and hence, the demonstrated
multiplication doesn't appear quite as startling. (17 and
19 are the next two such numbers.)
Hempel's Paradox
Part of the discipline, logic, represents classes of objects by suitable symbols. For example, the class A might
represent the class of all four-footed objects: chairs, tables,
dogs, etc. The logical statement that all A is B is equivalent to the statement that all not-B is not-A. This is not
very surprising if we let A represent the class of all normal
dogs, and let B represent the class of four-footed animals.
Then all A is B says that all normal dogs are four-footed
animals. And the phrase "all not-B is not-A" becomes
"All not-four-footed animals are not-normal dogs."
Bits and Pieces
33
The logical statement seems to assume a ridiculous

meaning if we take the sentence" All crows are black"
and transform it into "All not-black objects are notcrows." If one were to try to prove this statement experimentally, then he could say, "Here is a red pencil," and
that is, indeed, a not-crow. And "Here is a blue book,"
and it, too, is a not-crow. And" Here is a white cup," and
that, also, is a not-crow. It begins to look as if all crows
are black. Now this line of reasoning appears to be preposterous and useless.
But suppose we wish to investigate the truth of the
statement that all red-headed secretaries at a particular
company are married, and assume, further, that this investigation is to be done at a place where there are one
thousand secretaries, three hundred of whom have red
hair, and one hundred of whom are not married. We can
assign the class of red-haired secretaries to A and the
class of married secretaries to B. To investigate the statement that all red-headed secretaries are married, we would
have to ask all three hundred red-headed secretaries if
they were married-three hundred questions. On the other
hand, if we were to investigate the equivalent statement
" All not-married secretaries are not red-haired," then we
have only one hundred people to question. It is obviously
more economical to use the inverted form of the logical
statement even though it appeared ridiculous in the blackcrow example. This seeming inconsistency is known as
Hempel's Paradox.
Geometric Proof That
~m3 = [~m
The following proof was brought to my attention by

M. W. Green, who attributes it to S. Golomb. Discovery
34
Bits and Pieces
of this proof requires the ability to visualize the geometrical

construction shown in figure 13. In this figure, in the upper
left corner, there is a unit square which represents the
area 1~. To the right of that square is square 2~, and
below it is square 2i. To the right of square 2~ is square
3~. Below square 2i is square 3~, and along the diagonal
is square 3i. These three squares of size 3 2 each, have a
combined area of 3 X 3 2 or 33 We can continue drawing
the squares as \ve have done until now, and the drawing
will resemble the drawing shown in figure 13.
12
~A
2 ~
28
3 A2
3 2C
3 82
r--4 A2
52
42
4 20
52
~
~
2
4B
52
52
52
1
Figure 13
All the squares fit together very nicely except for the
even squares along the diagonal. Squares 2! and 2i oyerlap
Bits and Pieces
35
along the diagonal. But the area of overlap, which is shaded

on the drawing, is equal to the vacant area next to it
along the diagonal. Because of this beautiful interleaving,
we can write that the total area of the squares is:
Area = 1(12) + 2(22) + 3(3 2) + 4(42) ...
= 13 + 23 + 33 + 43 + ...
This area is also equal to [1 + 2 + 3 + 4 + ...)2. Since
the two areas are equal, this completes the demonstration
and none but mathematical purists would deny that it is
an effective and beautiful demonstration.
In deference to the purist, it must be recognized that
there are dangers inherent in too easy a generalization.
Some of the people who refused to accept the "obvious"
statement that parallel lines never meet must be given
credit for the development of non-Euclidean geometries.
But it is difficult to determine when the questioning of an
obvious statement is profitable and when it is not.
These thoughts bring to mind the story of the experimental physicist who set out to test the conjecture that
all odd numbers are prime. He started by saying, "Threethat's prime. Five-that's prime, too. Seven-still prime.
Nine-oops!-What's wrong? But let's go on a bit further.
Eleven-that's prime. Thirteen-still prime. Nine must
have been an experimental error!"
A Peculiar Infinite Sequence
Consider the infinite array
s = x:.:!'cP
For what values of x is S finite? People unfamiliar
with this problem are usually tempted to say that S
diverges for all values of x greater than one, and approaches
zero for all x less than 1. But such is not the case. It can
be shown easily, by means of a "log-log" slide rule, or by
36
Bits and Pieces
other means well-known to the mathematician, that

1.21.2 = 1.244, and 1.21.244 = 1.254. (Please note that for
any value of x, S must be evaluated from the top down.)
In fact, S = 1.258 when x = 1.2. And we can find finite
values of S for x < 1. For example, if x = .5, S = .641.
Table 7 gives the value of S for representative values of
x in the range 0 < x < 1.445.
x
.0001
.001
.01
.1
.3
.5
.7
.184
.22
.277
.40
.53
.641
.762
x
.9
1.0
1.1
1.2
1.3
1.4
1.445
.909
1.0
1.112
1.258
1.47
1.88
2.7183
Table 7
The highest value of x in this table is the eth root of e

(i.e., el/e), and this corresponds to S = e(2.7183). S does
diverge for x > ells.
We can approach the problem analytically in the
following way. Since S = x~cP, we can raise x to the S
power and get the equation X S = S, since one more x in the
infinite sequence does not make any difference. Taking logarithms of both sides, S log x = log S, or log x = (log S) / s.
But now a curious thing occurs. If x is only slightly larger
than 1, its log is very small, and log x can be small either
because log S is small, or because S is large. For each value
of 1 < x < ells in table 7 a second value of S can be found,
as is shown in table 8.
x-
1.1
1.2
1.3
1.4
s-
38.2
14.7
7.85
4.4
Table 8
Bits and Pieces
37
This second value of 8 can only be computed by

considering log x = (log 8)/8. The smaller value can be
computed directly by raising x to the appropriate power.
Wilkes-Gordon Construction
A farmer set out to erect a fence in a perfectly straight

line. The only tool he had to help him locate the postholes
was a straightedge. With it, he could continue any straight
line indefinitely. He started to layout the fence and suddenly found that the line ran smack into a large tree. He
didn't want to start the line over again to avoid the tree,
and he wanted to continue the line on the other side of
the tree. How could he do this, using only a straightedge?
The solution of the problem can be obtained by the
use of Desargues' Theorem, * or by a construction devised
by R. Wilkes and proved by G. Gordon.t Desargues'
Theorem is easy to prove in three dimensions, but difficult
to prove in two dimensions. The two-dimensional theorem
is required in this case.
OBSTRUCTION
Figure 14
* Courant and Robbins, What Is Mathematics'! New York:

Oxford Univ. Press, 1941, p. 170.
t Ibid., p. 197, problem 6, by Steiner, which is substantially the
same as the Wilkes-Gordon Construction.
38
Bits and Pieces
Desargues' Theorem in two dimensions can best be

described by reference to figure 14. It states that if two
triangles are drawn in a plane so that the three lines
passing through pairs of corresponding vertices intersect
at one point, then the extensions of corresponding sides
of the triangles will intersect in points that are collinear.
In figure 14 lines AD, BE, and CF intersect at a common
point, Pl. The theorem then states that the intersections
of BC with EF, AC with DF, and AB with DE, all intersect at points R, S, and T, which lie in a straight line. The
problem can also be solved by the Wilkes-Gordon* Construction. The Wilkes-Gordon Construction can be described with the aid of figure 15.
Figure 15
This construction states that if two lines from a common point, PI, are intersected by a fan of lines from a
* Both Mr. Wilkes and Dr. Gordon are engineers at RCA.
Bits and Pieces
39
second point, 0, then the intersections of the diagonals

of the quadrilaterals formed by these two sets of lines
(R, S, and T) are all collinear with Pl.
To solve this problem using Desargues' Theorem, we
refer to figure 14, and we choose Rand S to be on the line
before the obstruction. We choose a point PI arbitrarily,
and draw lines P 1P 2, l\P a, and P 1P 4 From point R, lines
RC and RF are drawn. From point S, lines SC and SF
are drawn. These lines define points B, A, D, and E,
as is shown in figure 14. Lines AB and ED intersect at a
point T, on the same straight line as Rand S. Then, Rl
and S] can be chosen (to be different from Rand S, but
on the same straight line). The entire procedure is repeated,
and point T] is defined. T and Tl are each collinear with
Rand S, and therefore define the line on the other side
of the obstruction.
The Wilkes-Gordon Construction can be used as
shown in figure 15. Point p] is chosen on the line which
had been drawn. From point PI, lines P 1P 2 and P 1P 4 are
drawn. Then a diagonal is drawn through the arbitrary
point R. (R is arbitrary except that it must be on the
original line.) From the intersection of that diagonal with
P]P4, a straight line is drawn. This straight line is labeled
3 in figure 15. Through point R and the intersection of
line 3 with p]P2, a second diagonal is drawn. The intersections of the two diagonals with P]P2 and P 1P 4 define
line 5. The intersection of line 3 with line 5 defines point
O. Now, all lines from 0 which intersect P]P2 and P]P4
define quadrilaterals, and the intersections of the diagonals
of these quadrilaterals (such as Sand T) are collinear
with PI and R.
Desargues' Theorem requires a new construction for
each individual point. The Wilkes-Gordon Construction
gives many points from one construction and is somewhat
more economical.
40
Bits and Pieces
Proof of the Pythagorean Theorem
The following proof of the Pythagorean Theorem has

a certain attractiveness about it, primarily, I suppose,
because it is so simple and candid. Consider the triangle
to be investigated with sides a, b, and c.
Figure 16
Let us now arrange four such triangles as shown in

figure 17. Since angles A and B are complementary, eaeh
Figure 17
Bits and Pieces
41
angle of the quadrilateral must be 90. And since each

side of the quadrilateral is c, the quadrilateral is a square.
According to one reckoning, the entire area will be (a + b)2,
and according to another reckoning it will be c2 and four
triangles. The area of each triangle is ab/2 and the expression for the total area reads:
(a
a2
+ b)2 = 4(ab/2) + c
+ 2ab + b2 =
a 2 + b2 =
2ab
c2
+ c2
The Advantages of Compound Interest

In order to attract customers, banks often offer interest, compounded not once or twice, but four times a year.
It requires but a few moments to calculate the actual
benefits of such compound interest. If the bank offers
interest at i%, compounded once a year, then the money
on deposit grows by a factor (1 + i) each year. If the
bank offers the i% interest, compounded n times a year,
then we must divide the year into n parts, and apply a
rate of interest (i/n) for each of those periods. Thus, the
money would grow by a factor [1 + (i/n)] each period or
[1 + (i/n)]" each year, because each year has n periods.
That is, the money invested at 3% compounded quarterly
is worth, at the end of the year, [1 + (.03/4)]4 times as
much. In table 9 are listed the equivalent factors for different intervals of compounding the interest. Also, shown
in this table are the factors applicable when the interest
is compounded continuously.
The factor by which money grows if the interest has
been compounded n times a year has been given as
[1 + (i/n)]n. As the interest is compounded more and
more frequently, n approaches infinity, and the factor
approaches the limit ei The expansion of ei is:
'2'3
, 1 +z.+ -+,
~
~ +
e'=
2!
3.
'n
~ + ...
... +,
n.
42
Bits and Pieces
Table 9 lists the equivalent factors for several rates

and different periods for compounding.
It can be seen from this table that the advantage of
having the interest compounded quarterly may, in certain
cases, be canceled by the cost of an additional postage
stamp.
n
2%
8%
4%
5%
1
2
4
1.02000
1.02010
1.02015
1.02020
1.03000
1.03023
1.03034
1.03045
1.04000
1.04040
1.04060
1.04081
1.05000
1.05063
1.05094
1.05127
00
Table 9
5: Coin Weighing Problems
Bachet's weight problem asks what set of weights is required to weigh all integral values of pounds (i.e., 1, 2, 3,
etc.) up to 31, if the weights can be placed on one side of
the balance only. A simple extension asks for those weights
required to weigh from 1 to 40, if one is allowed to put
weights on either side of the balance. In the first case the
answer is the progression of powers of 2 (the binary sequence), namely, 1,2,4, and 8, etc., and in the second case,
a progression of powers of 3, namely, 1, 3, 9, and 27, etc.
Another modification asks what weights are required to
weigh up to 12 pounds, if one has a four-pan balance. The
pans are symmetrical on opposite sides of the fulcrum, and
the outer pans have twice the lever arm of the inner ones.
The answers in this case are 1 and 5. (The general term
is 3".)
Determination of Odd Coin Among Twelve Coins
A more sophisticated problem is the following: You

are given a pile of twelve coins, all of which look the same,
and are of the same denomination, and are told that one,
and only one, of these coins is counterfeit. But you do not
know whether the counterfeit coin is heavier or lighter
than a true coin. Using an equal-arm balance, you are
43
44
Coin Weighing Problems
asked to identify the false coin in three weighings, and to

state whether it is heavier or lighter than the rest.
A solution to this problem requires the recognition
of two facts: (a) if equal numbers of coins are placed in the
two pans, and they balance, then the false coin must be in
the group of coins which have not yet been used, and (b) if
two groups of coins do not balance, then each coin in the
pan that goes down is potentially heavy (i.e., if the false
coin is heavy, it must be one of those, and each coin in
the pan that goes up is potentially light.)
It is convenient to label potentially heavy coins H, and
potentially light coins L. The original pile of 12 coins is
divided into 3 groups of four each, and two of these piles
are compared. If the two piles balance, we know the false
coin is in the last pile of four. Means for selecting the false
coin from those four will become apparent in the following
discussion. If the two original piles did not balance, each
of the coins in the heavier pan is labeled H, and each one
in the lighter pan is labeled L. The four coins in the last
group are labeled G (for good) because there is only one
false coin, and the last pile must accordingly be all good.
Then we put 2 Hand 2 L in the left pan for the second
weighing, and we put 1 L, 1 H, and 2 G in the right pan. If
the left pan goes down, then either one of the 2 H in the
left pan was actually heavy or the L in the right pan was
actually light. Then, weighing one of the 2 H and the L
against 2 G identifies the culprit. If they balance, the remaining H must be the false coin. If they do not, the H
or the L is at fault and can be weighed against the G.
If the first weighing balanced, we would then take
two of the remaining four, and weigh them against one of
those remaining four plus a good coin (taken from the
eight which must be all good). Again, 2 Hand 1 L (or its
image) are determined, or a single unknown coin is left.
In general, (3 N - 3) /2 coins (where N is the number of
weighings) can be weighed in this way.
45
Weak and Strong Solutions

The previous solution is known as a weak solution
because each arrangement of coins to be weighed depends
on the result of the previous weighing. "Strong" solutions
are those where the arrangements to be weighed are
specified in advance. (The words weak and strong are not
intended to reflect on the intellectual prowess of the problem solver. They are simply a convenient definition.) It is
quite reasonable that a strong solution for this problem
should exist. There are three possibilities for each weighing-the left pan can go down (8), the right pan can go
down (D), or they can balance (B). Since three weighings
are performed, there are 33 or 27 different possibilities.
One of these, the one where the pans balance each time,
gives us no information on whether the suspected coin is
heavy or light, and it must be discarded. The remaining
twenty-six possibilities can be arranged into thirteen
pairs, in which each arrangement of D's and 8's and B's
has its mirror image (where the D's and 8's are interchanged). The use of only one of each of these symmetrical
pairs enables us to determine whether it is this coin that
is heavy, or that coin that is light.
A Strong Solution
One of the easier ways to construct a strong solution
is to start by preparing a list of all groups of 3 D's, 8's, or
1
2
3
4
S
S
S
S
D B
5 D B S
6 B S D
7 D B B
8
9
10
11
12
13
Table 10
B
B
D B
B D
S D D
D S D
D D S
46
B's in which no group is the mirror image of any other.

Such a listing is shown in table 10.
From table 10, we can go on to derive the strong
solution as is shown in table 11:
1st weighing:
2nd weighing:
3rd weighing:
Left Pan (8)
Right Pan (D)
Not Used (B)
1,2,4,10
1,3,6,11
2,3,5, 12
5,7,11,12
4, 8, 10, 12
6,9,10,11
3, 6, 8, 9
2,5,7,9
1,4,7,8
Table 11
This table is constructed from table 10, by assuming

that each coin in turn is heavy, and assigning it to the left,
right, or unused pile for each of the three weighings in accordance with table 10. The possibility of its being light is
automatically provided for, because no mirror images
were used. A restriction, of course, is that the number of
coins on the left pan must equal that on the right pan. This
restriction prevents us from using the thirteenth possibility in table 10, namely SSS. Any other one of the thirteen listings could have been omitted instead of that one.
The odd coin is determined by finding that line of table 10
which corresponds to the actual happenings in the three
successive weighings.
The problem of the twelve coins can be modified to
use the thirteenth possibility, as shown in table 12. To do
this, consider a group of fourteen coins in which one coin
is pointed out as a true coin, and in which a false coin is
somewhere in the remaining thirteen. It is not known
1st weighing
2nd weighing
3rd weighing
Left Pan
Right Pan
Not Used
1,2,4, 10, 13
1, 3, 6, 11, 13
2,3,5, 12, 13
5,7, 11, 12, G

4, 8, 10, 12, G
6,9, 10, 11, G
3,6,8,9
2,5,7,9
1,4,7,8
Table 12
47
whether the false coin is heavier or lighter than the rest.

Table 12 shows a strong solution in which the thirteen
coins are numbered, and the fourteenth is labeled G.
This procedure will identify the false coin among
thirteen coins, and will indicate whether it is heavy or
light. A weak solution is also applicable to this problem.
Strong solutions are more difficult to prepare than
weak solutions, and this accounts for the fact that such
solutions are far less common. A weak solution for the
problem above with 120 coins and five weighings can
easily be done. A strong solution for this case is long and
arduous. In general, the odd coin from among (3 n - 3) /2
coins can be determined in n weighings, and it can be
determined whether that coin is heavy or light.
It is reported by Gnedenko, * that the Russian Mathematical Olympiads asked for the determination in four
weighings, of a single light coin among 80 coins. If it is
known that the coin is light (or heavy) then one of 3N
coins can be determined in N weighings. In the present
case, using four weighings, one out of 3 4 = 81 coins can
be determined. It may be that the number 80 (rather than
81) was used because 80 is not divisible by 3, and may be
pedagogically more appealing.
Another Coin Problem
Part of the definition of an interesting problem is
that it seems offhand to be impossible to solve. Accordingly, the following is a very interesting problem.
In a certain country, there are N (an arbitrary number) denominations of coins in use. Each denomination of
coin is made by one and only one company and the proper
weight of each kind of coin is specified by the legislature.
* Gnedenko, B. V., "Mathematical Education in the USSR,"

American Mathematical Monthly, 64 p. 389 (1957), Problem 3, p. 397.
48
One of the companies is suspected of cheating, but it is

not known whether it cheats by making the coins somewhat light (and pocketing the difference) or by using a
heavy base metal in the center of the coin, thus making
it heavier than it should be. Armed only with a list of the
proper weights and a spring balance, and having available
as many of each coin as we wish, determine in two weighings which company is cheating and determine the discrepancy. (It is assumed that only one company is cheating.)
This problem, at first glance, seems impossible. At
least one person prepared a "proof" of impossibility when
the problem was first proposed to him. Needless to say,
there was a flaw in his "proof." To solve this problem.
weigh a group consisting of one coin of each type. Knowing this total weight and also the sum of the legal weights,
determine the discrepancy, of the suspected coin. You still
won't know which coin is off by this amount. For the next
weighing, we take one of the first coin, two of the second,
three of the third, etc., and again find the discrepamy.
The discrepancy this time will be md where d is the previously determined mismatch, and m is the rank of the coin.
6: Fermat's Theorem; Sailors and Coconuts
One of the leading figures of classical number theory was

a Frenchman, Pierre Fermat. Fermat was a magistrate
most of his life, but nonetheless he made some of the
weightiest contributions to number theory. His theorems
can be proved using nothing more complicated than highschool algebra. Many people feel that Fermat's proofs
may have been done the same way. This statement is more
interesting in view of the following story:
Fermat was in the habit of merely announcing certain
theorems without presenting a proof. All but one of his
theorems have since been verified. The big exception is
his statement that 22" + 1 is always a prime number (has
no factors besides unity and itself), but he was unable to
prove it. Euler proved the statement false by showing that
2 25 + 1 is divisible by 641. (In fact, no primes have been
discovered among the Fermat numbers, 2211 + 1, for any n
greater than 4.)
How, then, are we to regard Fermat's notation on the
margin of his copy of Diophantus that he had found "a
most elegant proof of the statement that xn + yn = zn
has no solution in integers for n ~ 3. But this margin is
too small to contain the proof." Mathematicians have been
battling with this problem ever since, and have still not
found a proof.
The Fermat "blooper" raises a very interesting
49
50
Fermat's Theorem; Sailors and Coconuts
question, namely, how can one determine whether a number is prime. Fermat, as well as other number theorists,
was deeply concerned with this problem. The mathematicians of Fermat's time may have had a slight advantage because they apparently had wonderful ways of
recognizing prime numbers. Such recognition seems
uncanny today.
For example, Fermat was asked whether or not the
number 100,895,598,169 was prime. Fermat is said to have
answered instantly that the number is not prime, but that
it is the product of 898,423 and 112,303, and further that
both of these were prime.
The field of number theory interests not only the hard
core of number theorists, but also a soft outer shell of
gifted amateurs. Wilson was one such amateur. He derived
an expression for n which, if satisfied, is both necessary
and sufficient for n to be a prime number. Wilson's theorem
states that if (n - 1)!* + 1 is evenly divisible by n, then,
and only then, is n a prime number. This is certainly a
beautifully simple condition for necessity and sufficiency,
but the difficulty in calculating (n - I)!, even when n is
small, limits the usefulness of the theorem.
In order to understand the Fermat theorem with
which this chapter is concerned, we must become familiar
with two new notions. The first is the notion of modulus.
In using this concept, we say that two numbers are congruent modulo N if the remainders, upon dividing each
number by N, are equal. For example, 7 == 2 (mod 5)
[read: 7 is congruent to 2 (modulo 5)], or 11 == 2 (mod 3).
(This simply means that both 7 and 2, when divided by 5,
have the same remainder-2. Likewise, 11 and 2 when
divided by 3 both have a remainder of 2.) We use this
technique every day when we speak of the time of the day
or the day of the week, since we subconsciously subtract an
* (n
- I)! (where n = 6) would be 5! or 5 X 4 X 3 X 2 X 1.
51
integral number of days (and hours) in the first case, and

an integral number of weeks in the second.
The second new idea is that of the Euler cp function:
cp(n) is defined as the number of integers (including 1),
which are smaller than n and are prime to n (i.e., have no
common factors with n). Thus, cp(6) = 2 because only 1
and 5 are less than 6 and prime to it. Also cp(7) = 6 and
cp(8) = 4. Fermat's theorem states that a",(n) == 1 (mod n)
if a is prime to n.
Considering that cp(n) appears to be such an unwieldy
function, it is surprising that such a simple relationship
exists. cp(3) = 2 because only 1 and 2 are less than 3, and
prime to it. Thus, Fermat's theorem states that:
a",(3)
== 1 (mod 3) or a 2 == 1 (mod 3)
This states that the square of any number which is not a

multiple of 3 has a remainder of one when divided by 3.
E.g., 52 = 25 == 1 (mod 3), 72 = 49 == 1 (mod 3). Please
note, however, that 62 = 36 == 0 (mod 3). The restriction
that a be not divisible by n is quite apparent. Similarly, the
fourth power of any number not divisible by 5 has a remainder 1 when divided by 5. Thus, 24 = 16 == 1 (mod 5),
3 4 = 81 == 1 (mod 5). But 54 = 625 ~ 0 (mod 5). A proof
of this theorem is given at the end of this chapter.
Fermat's theorem is the key to the following problem:
What is the smallest number such that moving the righthand digit to the front of the number (and shifting all the
other digits one place to the right) merely doubles the
original number? This problem can be attacked by the
"hammer and tongs" approach, in which we say the end
number must be at least 2 because the original number
must start with at least 1. Then, working to the left, the
successive numbers must be 4, then 8, then 6 (but carry
one), then 3 (carry one), etc. We stop when we come to
a zero or a 1, preceded by a number less than 5. (If it were
greater than 4, double the number would start with a 3.)
52
But there is little doubt that this method is hardly

elegant, and surely possesses no generality. A much more
satisfying solution is the following one:
Let the number we seek, N, be represented by:
N
anIOn
+ an_llOn- + ... + adO + aolOo

1
(1)
We can also write:

2N = aolO n
+ anlOn- + ... + a 101 + allOo

1
(2)
Multiply (2) by 10, getting:

20N
aolO n+1
+ anIOn + an_llO 1 + ... + allOl

n-
Subtract (1) from (3):

19N = ao(lOn+1 - 1); N = ao (
lOn+1 -
19
(3)
1)
Fermat's theorem states that the factor in parentheses

is an integer when n + 1 = 18 (because tI>[19j = 18).
The hammer-and-tongs solution corresponds to the case
for ao = 2. The case for ao = 1 is a solution only if we
agree to accept the first digit in the quotient as really a O.
Thus: 1018 - 1 = 999999999999999999; (10 18 - 1)/19 is
shown in figure 18.
This number is a solution for all except the purist who
insists that zeros to the left of the first significant digit are
not part of the number and should be neglected. For him,
the solution is double this number.
The original question could have asked for the altered
number to be 3 times as large as the first number. In this
case, the number would have come out to be a multiple of
(lOn+1 - 1)/29. Fermat's theorem requires n + 1 = 28
in this case. Such large numbers, however, are cumbersome
and hardly suitable for chats about" nice little problems."
Sometimes the problem is varied so that moving the first
digit to the right side increases the original number by
fifty percent.
052631578947368421
19)999999999999999999
95
49
38
119
114
59
57
29
19
109
95
149
133
169
152
179
171
89
76
139
133
69
57
129
114
159
152
79
76
39
38
19
19
o
Figure 18
53
54
Three Men and Twelve Lumps of Sugar

The study of number theory can be fas('inating. Its
application ranges from exceedingly simple subjects to
extremely abstruse ones. At the complicated end of the
spectrum is Kummer's discovery that no integer n which
divides a llumerator of a specific set of Bernoulli numbers*
can possibly satisfy Fermat's equation xn + yn = zn in
integers. The other end of the scale is indicated by the
following story:
Three men went into a diner, and each ordered a cup
of coffee. The waitress brought the three cups of coffee
and a dish with twelve lumps of sugar. Each man took an
odd number of lumps of sugar, and when they had finished,
there was no sugar left. How many lumps did each man
take?
It requires only a few moments to recognize that the
sum of three odd numbers must be odd itself. So there
must be a trick somewhere, and there is.
The first man took one lump, the second man took
one lump, and the third man took ten lumps. "Aha!" you
will cry, "ten is not an odd number!" And then, we slyly
inquire, "Do you know anyone who takes ten lumps of
sugar in his coffee?"
Perhaps a better example of simple number theory is
the problem asking for a proof that all primes are of the
form 6k 1. "Forsooth!" the reader may say. "This
chap has, indeed, lost his marbles. There is no formula to
give only primes." Please note that it states that all primes
must be of this form-not that all numbers of this form are
prIme.
To prove this, we note that 6k + 2 can be factored
into 2(3k + 1), 6k + 3 into 3(2k + 1), 6k + 4 into
* The Bernoulli numbers Bn are those which satisfy the expression

co
xn
co
Bnxn
L
XL-=1
n=o(n + I)!
n=O n!
55
2(3k + 2), and 6k + 5 is equivalent to 6k' - 1 (where k'

is simply one larger than Ie). So all numbers which are not
6k 1 are composite, and the primes, then, must exist
among the 6k 1.
Sailors and Coconuts
We now move on to the problem of the five shipwrecked sailors who collect a pile of coconuts and decide to
divide it the following day. In the middle of the night, one
of the sailors wakes up. To make sure he isn't cheated the
next day, he takes his fifth of the coconuts right then. He
counts them, and finding one too many for the total number to be divisible by five, throws it away, takes his share,
and goes back to sleep. Each of the others goes through the
same process, sequentially, of waking up, finding one too
many for the number to be divisible by five, throwing the
extra away, and taking his fifth. What is the smallest
number of coconuts that the original pile might have
contained? The modulus technique can be used to solve
this problem but it is fairly cumbersome. Actually, the
following solution, attributed to E. Fermi, * has more charm
and appeal.
This solution says minus four is surely a mathematical
solution, because each man, seeing minus four coconuts,
throws one away, leaving minus five, and takes his fifthleaving minus four again. To get a real and positive solution, however, we must add the smallest number divisible
by 5, 5 times, which is 55. So the answer is:
55 - 4 = 3,125 - 4 = 3,121
An interesting note might be added here. Several

years ago, a problem appeared in the" Letters to the Editor" column of Life magazine. This problem asked for
* The truth of this story is immaterial. It indicates the high

regard that the physics community had for Fermi.
56
the solution of the aforementioned problem for three

sailors, each throwing one coconut away before taking his
third, but with the added condition that the next morning,
a fourth coconut had to be thrown away in order to divide
the pile into thirds. The solution to this problem is "obviously" 3 4 - 2, or 79, if we are aware of the solution attributed to Fermi. Two weeks after the problem appeared all
sorts of people had written in their answers. A banker said
he did it in two nights, a lawyer had done it in three hours,
etc. A mathematics student from M.LT. bragged that he
had done it in twelve minutes! The answer can be written
down instantly-if we only know how!
The proof that nn - (n - 1) is a solution for the
problem of the sailors and coconuts is fairly straightforward. We start with
[nn - (n - 1)] coconuts
After throwing one away, and taking one nth, there are left:
n-1
- - [nn - n] = [(n - l)n(n-l) - (n - 1)]
And after the second person throws one away, and takes
an nth, there are left:
n-1
- - [(n - l)n(n-l) - n)]
= (n - 1)2n(n-2) - (n - 1)
This process can not be continued after the nth removal of

one nth of the objects, because the term which started out
as nn will have been metamorphosed into (n - l)nno. If it
were desired that the process be performed r times, we
would have had to start with a number of objects
n r - (n - 1).
In all the versions of this problem that I have seen,
the number of people has been prime. This restriction,
however, is required only if we wish the number of coco-
57
nuts, on the next morning to be evenly divisible by the

number of people. (Fermat's theorem tells us that
(n - 1)" - (n - 1) = (n - 1)[(n - 1),,-1 - 1] is always
divisible by n, if n is a prime number.)
For example, if the problem of the sailors and coconuts
involved only four sailors, each taking his fourth (after
throwing one away), the number of coconuts to start with
would have been 44 - 3 = 256 - 3 = 253.
Table 13 gives values of N (the total number of objects) for different values of n (the total number of people)
and shows the amount left by each person at each step in
the process. It is seen that the number of objects left by
the last person is divisible by the total number of people
only when n is a prime.
Total
number
1st
of
person
n coconuts leaves
2
3
4
5
6
3
25
253
3,121
46,651
2nd
person
leaves
3rd
person
leaves
4th
person
leaves
5th
person
leaves
6th
person
leaves
1
0
16
10
6
189
141
105
78
2,496
1,996 1,596
1,276 1,020
38,875 32,395 26,995 22,495 18,745 15,620
Table 13
a5
a Is Divisible by 30
Another problem which yields very neatly to Fermat's

theorem asks for a proof that all numbers of the form
a5 - a are evenly divisible by 30. To prove this, we first
factor an a out of the above expression, getting a(a 4 - 1).
Fermat's theorem states that a 4 - 1 is always divisible by
5, unless a itself is a multiple of 5. So a(a 4 - 1) is always
divisible by 5. Then a 4 - 1 is factored, and the original
number becomes a(a 2 + l)(a 2 - 1). By an argument
58
similar to the above, a(a 2 - 1) is always divisible by 3.

Finally a 2 - 1 is factored into (a + l)(a - 1), and one
of the two numbers a or a + 1 must be divisible by 2. We
have thus shown that (a 5 - a) is always divisible by 2, 3,
and 5, the product of which is 30.
Divisibility of a 13
A problem which is similar to the above, but which is

more extensive, is the one asking for a proof that 2,730 is
a factor of any al3 - a. The prime factors of 2,730 are
2, 3, 5, 7, and 13. We must show divisibility by each of
these numbers.
Fermat's theorem tells us that 13 is a factor of any
a l3 - a. a l3 - a can be factored into a(a6 + 1)(a6 - 1).
Seven is a factor of any a(a6 - 1). The (a 6 - 1) can be
factored into (a 3 + 1)(a3 - 1). The (a 3 + 1) can be
factored into (a + 1)(a 2 - a + 1) and the (a3 - 1) into
(a - 1)(a 2 + a + 1). The entire number has been factored
thus:
(a 13
a) =
a(a6
+ l)(a + 1)(a
2 -
+ l)(a -
1)(a 2
+ a + 1)
One of three numbers in a row is divisible by 3, and one of

two numbers in a row is divisible by 2. Therefore, the
product (a - l)a(a + 1) must be divisible by 2 and 3.
We have shown divisibility by 2, 3, 7, and 13, and only 5
remains to be shown as a divisor.
We must now change tactics and write a as (10k + r).
(This is true for any a we choose.) If the r happens to be 0
or 5, a is surely divisible by 5. If r is 1 or 6, then the (a - 1)
term would be divisible by 5, and similarly, if r is 4 or 9,
the (a + 1) term is divisible by 5. If r is equal to 2, then
the (a 6 + 1) would be divisible by 5-because each member of the expansion of (10k + 2)6 + 1 is divisible by 10,
except for the term 26 + 1 = 64 + 1 = 65, which is again
divisible by 5. By a similar argument, if r = 3 or 7 or 8,
59
each number in the expansion would be divisible by 10

(or 5), except for the 3 or 7 or 8 to the sixth power plus 1.
But 3 or 7 or 8 to the sixth power always ends in a 4 or a 9,
and therefore (a 6 + 1) would again be divisible by 5. The
known factors of a 13 - a now include 5, and are thus 13,
7, 5, 3, and 2, and the product of all these factors is 2,730.
Proof of Fermat's Theorem
Fermat's theorem states that at/>(n) == 1 (mod n)
where a is prime to nand cp(n) has been defined as the
number of numbers smaller than n and prime to it. Let
these numbers be represented by:
There can be only cp(n) different ones. Multiply each of

these numbers by a and write each of these products as
an integral number of n's plus some remainder, R. Thus:
aSl
Qln
+R
1;
aS2
Q2n
R2;
aSt/>(n)
Qt/>(n)n
+ Rt/>(n)
(1)
There are cp(n) such equations because there are cp(n) different s's. The Q's above are not necessarily different from
each other. Some may even be zero. Each of the R's must
be prime to n. If this were not so, we could factor out the
common factor from it and n on the right, and either a or 8
would then have to be divisible by this factor of n. But
this is contrary to our assumptions. Each R must also be
different from any other R. For if any two R's-say, R"
and R1/ - were equal, we could subtract those two equations from each other and get:
as" - aS1/
(Q"n - Q1/n)
+ R", -
Since R", = R7I we can factor and write:
R1/
60
The Fundamental Theorem of arithmetic tells us that if n

is a factor on the right side of the equation, then it must
also be a factor on the left side. But all s's are less than n
and prime to it, and a is prime to n. So, n cannot be a factor
of the left side of the equation. This impasse can only
mean that no Rz = R 7I Since the R's are all smaller than
n, and prime to n, and different from each other, and there
are cp(n) of them, they must be all the s's in some different
order.
Let us now multiply all the equations in (1) together.
We get:
aop(7O)[S1
S2
S3
X ... X
sop(7o)] =
+ [R
Ln
R2
R3
X ... X
Rop(7O)]
L represents some large number whose actual form does

not interest us. The product in square brackets on the
right must be the same as that in square brackets on the
left. We can thus factor and write
(aop(7O) -
1)[S1
S2
X ... X
sop(7o)] =
Ln
Since all the s's are prime to n, their product is also, and
it must then be aop(7O) - 1 which is divisible by n. Thus:
aop(n)
== 1 (mod n)
7: Partitioning Problems*
The Maid at the Stream
A maid is sent to a stream to get exactly two quarts

of water, but she has only a 7-quart jug and an ll-quart
jug. How can she measure the desired amount?
The Euclidean Algorithm, which is described at the
end of this chapter, yields integers rand s which satisfy
the equation:
rp-sq=l
(1)
where p and q have no common factor, i.e., are relatively
prime integers. For the purposes of this problem, this
equation can be interpreted as saying that if a p-unit container is filled exactly r times, and is emptied only into a
q-unit container, and this q-unit container is emptied
exactly s times (into the master container), then exactly
one unit must remain in the p-unit container.
If the desired amount were not a single unit, but a
units of the substance, then each term in equation (1)
could be multiplied by a, giving: arp - asq = a.
This equation indicates a way of obtaining a units,
but it might not indicate the procedure requiring the
minimum number of transfer operations. In such a case,
two things can be done. If, in the equation above, ar > q
*This chapter is an expansion of "A Solution for Certain Types
of Partitioning Problems" by M. H. Greenblatt, The Mathematics
Teacher, vol. 54, no. 7, p. 556.
61
62
Partitioning Problems
and as > p, then some multiple of pq can and should be

subtracted from each term on the left. This altered equation will then indicate a more economical solution of the
problem. Also, the equation:
r'p - s'q
(2)
which can be obtained by judicious manipulation of (1),

should be investigated. In this case, in which r' and s'
are different from rand s, the roles of the p and q containers are reversed; the q container is filled s' times and
the p container is emptied r' times. Again, in this case,
the terms on the left in equation (2) should be decreased
by a multiple of pq if possible. Either this solution or the
one derived from equation (1) will be the most economical
solution. Equation (1) and equation (2) may lead to the
same solution, and then both are equally economical.
These ideas can be clearly illustrated by solving an
actual problem. Suppose, for example, that a maid must
get exactly two quarts from a stream, and she has only a
7-quart jug and an 11-quart jug available to her. From
the Euclidean Algorithm we find (2 X 11) - (3 X 7) = 1.
As shown above, this can be changed to (4 X 11) 6
(6 X 7) = 2 and this equation states that at least 4
filling-and-emptying operations are required to get two
quarts if one starts by filling the II-quart jug. To investigate the alternate solution, we proceed as follows:
Multiply by 6:
Subtract 7:
Subtract 7 X 11:
Multiply by 2:
Subtract 7 X 11:
(2
(12
(12
(5
(10
(3
X
X
X
X
X
X
11)
11)
11)
11)
11)
11)
- (3 X 7) = 1
- (18 X 7) = 6
- (19 X 7) = 6 - 7 = -1
- (8 X 7) = -1
- (16 X 7) = -2
- (5 X 7) = -2
According to this solution, 3

5 filling and emptying
operations are required to get two quarts if one starts by
filling the 7-quart jug. This second solution is hence the
63
more economical one for this problem. The total number

of transfer operations is twice the number of filling and
emptying operations minus one. This is so because each
filling operation must be followed by a transfer to the
other container, and each emptying operation, except
the last, must be followed by a transfer from the other
container.
The question of whether the source is very large (as
in the case of the stream) or very small is not important
because the size of the source does not enter into the
solution. The only requirement is that the source capacity
be at least (p + q - 1). If the size of the source is less
than this, certain mathematical solutions are not physically realizable. If the numbers p and q are not relatively
prime, equation (1) is no longer true, certain numbers can
then not be formed by multiplication by a, and the problem
loses some of its interest.
Division of Balsam Among Three People
I once came across a variation of this problem in which
three men had stolen a jar containing 24 ounces of balsam
and wanted to divide it equally among themselves. They
found a 5-ounce container, a 9-ounce container, and a 10ounce container. (How improbable can these problems get,
what with balsam and such unusual size containers?) This
problem is subject to the same analysis as has been given,
but there is at least one pitfall which must be avoided. It
can, in a sense, be a much tougher problem than the one
we have considered.
Most people can see that 9 ounces can be put into
the lO-ounce container, and then the IO-ounce container
filled from the refilled 9-ounce container leaving the required 8 ounces in the 9-ounce container. But at this point,
the 8 ounces in the 9-ounce container must be transferred
to the lO-ounce container. Otherwise, since 5 and 10 are
64
not relatively prime, we will not be able to form the second

8 ounces. However, if the 8 ounces is put into the IO-ounce
container, then we have a 5-ounce and 9-ounce container
to operate with, and since 5 and 9 are relatively prime,
we can form the second 8-ounce portion. The steps are:
after pouring everything but the 8 ounces in the lO-ounce
container back into the master container, that is 16 ounces,
we fill the 9-ounce container, and from that fill the 5-ounce
container, leaving 4 ounces in the 9-ounce container.
These 4 ounces are put into the 5-ounce container, and
from the refilled 9-ounce container, only one ounce can
be poured into the 5-ounce container, which already has
4 ounces in it. Thus we have 8 ounces in the lO-ounce
container, 8 ounces in the 9-ounce container, and there
must then be 8 ounces between the 5-ounce container, and
the master container. The division is completed.
In another variation, the three men would have 5-,
7-, and ll-ounce containers, and since these three numbers
are all relatively prime, it can be solved by the methods
previously outlined.
Another Urn Problem
Another man had a 5-unit jug, a 9-unit jug and 14
units of a substance which he wanted to divide into two
equal portions. The Euclidean Algorithm tells us that:
7
(3
+4 =
7)
and that: (5 X 5) - (2 X 9) = 7
(5
+2=
7)
(3 X 9) - (4 X 5)
We can produce 7 units by filling either the 5-unit container, or the 9-unit container first. The total number of
operations in both cases will be:
(2 X 7) - 1 = 13
Additional problems of this sort can easily be devised.
65
Pile of Bricks
The same methods apply to solving the one about

the man who had a pile of bricks. The number of bricks
was such that when he tried to build a wall 2 bricks high
there was one left over. When he tried to build it 3 bricks
high, there was still one left over, and so on, for 4, 5, and
6 bricks in height. But when he tried to build the wall 7
bricks high, he had no bricks left over. What was the
minimum number of bricks in the pile?
We can solve this problem by simply trying each
number from seven on up until we finally discover that
301 has the desired properties. In addition to being timeconsuming and inglorious, this method possesses little generality. A much better solution is the following:
The least common multiple of 2, 3, 4, 5, and 6 (that
is, the smallest number evenly divisible by each of these)
is 60, where all the prime factors of each number appear
just once. Any multiple of 60, plus 1, will leave a remainder
of one when divided by 2, 3, 4, 5, or 6. As n is varied, the
numbers 60n + 1 will be congruent to each of the integers
from 0 to 6 (mod 7). We seek the lowest one congruent to 0
(mod 7), and in this case, (5 X 60)
1 == 0 (mod 7).
Again, 301 is the answer.
The multiple of 60 can be determined by trying each
integer from 1 to 7, or it can be derived from the Euclidean
Algorithm by finding the solution of the equation:
7r - 60s = 1
The number of bricks by which the pile is divisible

should be a prime number. Otherwise a solution may
not exist.
Poker Chips
A variation of this problem concerns a pile of poker

chips. When it is divided into piles of two chips each,
66
there is one chip left over. When it is divided into piles of

three, there are two left over, in piles of four, there are
three left over, and so on for five, six, seven, eight, nine,
and ten. But when the chips are divided into groups of
eleven, there is none left over. What is the minimum
number of chips in the pile?
As before, we form the number 2 X 3 X 2 X 5 X
7 X 2 X 3 = 2,520, which is the smallest number divisible
by each of the integers 2 through 10. All numbers of the
form 2,52On - 1 lack one unit to be divisible by the numbers 2 through 10. We must find that number of the family
which is divisible by 11. The first number we try happens to
have the desired property, because 2,519 + 11 equals 229.
Finding Any Rational Fraction
of a Volume V
The milkmaid in the first problem had a fairly difficult assignment, but consider the plight of the maid in
the present problem. She is given a glass jar of irregular
cross section and of volume V, an earthenware jug of
volume somewhat greater than V, a marking pencil, an
infinite source of water, and a cap for the glass jar. She
is required to produce any rational fraction of the volume
V. (Most milkmaids would quit at this point.)
The solution consists of two parts. First, we show
how the volume V can be divided into halves, quarters,
eighths, etc., and then we show how any rational fraction
can be represented in binary form (i.e., how many halves,
quarters, eighths, etc., must be added to equal any given
fraction).
First, we put in an amount of water which we think
is V /2. Then, with the marking pencil, we put a mark
at that level of the liquid, put the cap on the jar, and turn
it upside down. If the jar was half full of liquid, the level
of the liquid will be exactly at the mark that we made.
67
If there is too much liquid, the inverted level will be

above the mark; if there is too little liquid, the inverted
level will be below the mark. In either case, we can add
or subtract liquid and, by repeated trials, approach V /2
as closely as we wish. Finally, all the incorrect marks can
be erased, leaving only the correct one.
Next, the volume V /2 is poured into the earthenware
jug, and a volume which we guess to be V /4 is put into
the jar. Again, we put a mark at the guessed V /4 level,
add the V /2 from the jug, cap the jar, and turn it upside
down. The level now should be up to the V /4 mark. In
the same way as before, V/4 can be approached as closely
as we wish. This procedure can then be repeated with the
successive reciprocal powers of 2. Eventually, each fraction
of the form V /2 n and the sum of any of these fractions can
be marked.
Any rational fraction can be expressed as a repeating
decimal. This repeating decimal can be expressed in binary
form in the following way:
The decimal is written out and multiplied by 2. If
the product is less than one, then we write down a zero
as part of the binary representation. If the product is
greater than one we write a one for part of the binary
0.200
X2
@400 ...
~
@800 ...
X2
1600 ...
~
1200 ...
~
0400 ...
Figure 19
68
representation. Then the decimal part of the product is

multiplied by 2 again, and whenever the product in decimal
form is greater than one, we write a 1 for the binary representation, and simply neglect the one in the decimal for
subsequent mUltiplications. The procedure thus consists
of a successive multiplication of the decimal by 2, and the
assignation of a zero or a one for the binary expression.
Figure 19 shows how the fraction i can be expressed in
binary form. Having the binary expression, one can then
pour from the jar into the earthenware jug the required
fractions V /2n.
In a variation of this problem, the volume V in the
previous problem can be specified by a mark on the irregularly shaped jar, and a cap for the jar is not supplied.
Again, any rational fraction of V is requested. In this
case, the volume V of water is poured into the glass jar.
Then water is poured from the jar into the jug until we
guess that we have V /2 left. At this point, we mark the
jar, pour the water in the jar out, and pour the water from
the jug back to the jar. If the guessed volume were less
than V /2, then the level of water from the jug would be
higher and vice versa. If the levels were different, then
we would take as our next guess a level between the two
(for V /2 is greater than the lower mark, and less than
the upper mark).
To determine V /4, we pour from V (into the jug)
until we have what we think is V /4 left in the jar. One
marks this level of liquid. We can then empty this water,
pour V /2 from the jug into the jar (since V /2 has been
marked on the jar) and empty that, and put the remainder
from the jug into the jar. If the mark is higher than the
level of the water, the mark is more than V /4, and vice
versa. In any case, the various V /2 n and combinations
thereof can be marked. From this point on, the solution
is the same as the previous one. (The milkmaid has just
shot herself!)
69
The Euclidean Algorithm
The Euclidean Algorithm provides a simple way to

find the greatest common denominator (OeD) of any
two numbers. If the two numbers are relatively prime,
(p and q) then a very modest amount of additional work
provides the solution to rp - sq = 1. The Euclidean Algorithm tells us to write p as an integral number of q's plus
some remainder Rl-smaller than q. And then q is written
as an integral number of R1's plus some R2 (smaller than
RI). Then RI in terms of R2 plus some R3 R2) and so on,
until the remainder Rn is zero. The remainder just before
that, namely R n _ l , is the OeD of p and q. This method
is best illustrated by an example:
Find the OeD of 58 and 5. We write:
58
5 =
3 =
2
+3
(1 X 3) + 2
(1 X 2) + 1
(2 X 1) + 0
(11 X 5)
(3)
(4)
(5)
(6)
So 1 is the OeD of 58 and 5, and the two numbers

are relatively prime. To put these numbers into the equation required for the solution of this problem, we can
proceed as follows:
From (5) we can write:
= (1 X 3) - (1 X 2)
(7)
2 = (1 X 5) - (1 X 3)
(8)
From (4):
Substituting into (7):

1
= (1 X 3) - 1 (1 X 5 - 1 X 3)
=
(2 X 3) - (1 X 5)
(9)
70
And from (3):

3
58 - (11 X 5)
(10)
Substituting into (9):

1 = 2(58 - 11 X 5) - 1 X 5
= (2 X 58) - (23 X 5)
Q.E.I.
(Quod erat inveniendum)
8: Symmetry
Sam Loyd, who has been called America's Greatest Puzzlist, lived in the latter part of the nineteenth century and
the early part of the present century. He was a good chess
player in college, he later wrote a chess column for a New
York newspaper. He discovered that he also had a flair
for inventing puzzles and some of his puzzles are worldfamous. The 14-15 puzzle is a rather well-known example.
It consists of a square in which there are 15 smaller squares,
which can be moved around, and one empty square. The
15 squares are numbered 1 through 15. The question is
always to find whether or not the squares can be rearranged
in a desired order. So popular was this puzzle that people
neglected their regular duties to work on it. Shopkeepers
neglected their shops, and in France, the Legislature was
forced to ban the puzzle in order to get the legislators
back to work. Sam Loyd's life and puzzles have been
adequately described elsewhere. I include here Sam Loyd's
Daisy Puzzle because it is not well known.
Sam Loyd's Daisy Puzzle
This is a game between two people and is played

with a 13-petaled daisy. The players take turns picking
the petals, and they may pick either anyone petal or any
two right next to each other, if such a pair exists. The
71
72
Symmetry
person who picks the last petal wins. What is the proper
strategy for this game?
This game can be completely analyzed. The second
player can always win, and the proper strategy thus becomes a recipe-a straightforward procedure which guarantees a win. The first player must take either one or two
petals. The second player then takes two or one diametrically opposite. The first player then faces a daisy as is
shown in figure 20, regardless of his first move.
Figure 20
There are now two symmetrical groups of five petals

each, and whatever the first player does in one group, the
second player can do in the other. If the first player can
take a petal, so can the second player, and therefore the
second player always wins. It is quite apparent that the
essence of his success is symmetry.
Cigars on Table
Another game for two people requires a supply of

cigars and an ordinary table. The only restriction is that
the table have two mutually perpendicular axes of symmetry. The rules of the game require each of the two
Symmetry
73
players to alternately place a cigar anywhere on the table,

provided the cigar does not encroach upon any cigar already there. The first person who cannot legally place
another cigar on the table loses the game. This is another
game which is fairly simple to analyze, but in this case the
first person can always win.
The first person places the center of his cigar at the
center of the table, and the axis of the cigar is along one
of the table's axes of symmetry. Each position on one side
of this axis has a mirror image on the other side. Then,
wherever the second player places his cigar, the first player
puts his cigar in the mirror image of this position. So long
as the second player can put down a cigar, the first player
can put his in the mirror image. Thus, the first player
is guaranteed victory.
Cup of Coffee and Cup of Tea
A problem which can generate violent arguments concerns the mixing of a cup of coffee and a cup of tea. We are
given a cup of coffee and a cup of tea, containing equal
volumes. We take a teaspoon of coffee, put it in the tea
cup, and stir thoroughly. Then we take a teaspoonful of
the mixture and put it back in the coffee cup. Does the
coffee now have more tea in it than the tea has coffee,
or vice versa?
An easy way to see that each cup of liquid is equally
adulterated with liquid from the other cup is to forget
about the details of the transfer, and consider only the
final situation.
We started with equal volumes of liquid in each cup,
and we ended with equal volumes in each. At the end,
there is a certain amount of tea in the coffee cup. It must
have replaced an equal volume of coffee, which was in the
coffee cup at the beginning. And that missing coffee must
now be in the tea cup.
74
Symmetry
Some people are not convinced by this argument.

They point out that a full teaspoon of coffee was put in
the tea, but then some of that coffee was brought back
on the second transfer. And they say that there is then
less than 1 teaspoonful of coffee in the tea cup. But they
forget that because of this small amount of coffee coming
back, there is not room for a full teaspoon of pure tea to
mix in the coffee. A numerical example is very easy to
work out and may be surprising.
Nim
Another problem game which uses symmetry is the
game of Nim from the word "nim"-to filch. Nim is very
old and well understood. The game is for two people and
is played in the following manner:
Counters of some sort are arranged in N groups, each
group having an arbitrary number. Each player, in tum,
may take as many counters as he wishes from anyone
group. The player who takes the last counter wins (or
loses, depending on what was decided at the beginning
of the game). One of the more popular starting arrangements is shown in figure 21.
IIIII
IIII
III
101
100
11
212
Figure 21
The rules for winning this game usually mystify the

novice. The number of counters in each group is first
written as a binary number, as is shown in figure 21.
(Just as an ordinary, or decimal, number tells how many
of each successive power of 10 to accumulate, a binary
number tells how many of each successive power of two
to accumulate.) Each column of binary numbers is then
Symmetry
75
added up in the decimal system. The sum will be one of

two types. Each number in the sum will be even, or at
least one digit will be odd. If each number is even, we
will call it a "safe" situation, and if at least one number
in the sum is odd, we will call it an "unsafe" situation.
The rules for winning require the aspirant to leave
a safe situation to his opponent each time he moves. The
opponent, facing a safe situation, must necessarily leave
an unsafe one, because he must change at least one of
the binary digits, and he can do so in only one group.
The first player can then always change this unsafe total
into a safe situation by taking counters from only one,
carefully chosen, group. This can be seen by considering
the digits in the sum of an unsafe group. There will be a
left-most odd digit in the sum, and somewhere in the
binary representations there must be a 1 in this column
which corresponds to the odd digit. We choose to operate
in any of the groups which has a 1 in this column. We can
always remove a number of counters such that the offensive 1 is changed to a 0, and the binary digits to the right
of that 0 can be whatever we choose. And we choose them
so that the situation becomes safe. In this manner, the
first player keeps control over the game until finally he
faces a 2, 0 situation, or many groups of which only one
has more than one counter in it. This is as far as the rules
will take us, but it is far enough. The proper move in
either of these situations can easily be chosen.
There is a way of playing Nim which does not require
the binary notation and decimal addition to decide on the
right move. This method is actually the same as the
method which has just been described, but is somewhat
easier for mental calculation.
Inspection of the groups reveals that group or those
groups which have the highest power of two (a four or
an eight, etc.) in them. Counters corresponding to this
highest power of two must be arranged in pairs from
76
Symmetry
THESE GROUPS OF
I ARE PAIRED
2 GROUPS OF
8 EACH ARE
FIRST PAIRED
NO OTHER GROUP OF
4 TO MATCH THIS
SO WE MUST WORK
IN THIS GROUP
THE PROPER MOVE IS SEEN TO BE TO REMOVE THE

TWO MATCHES FROM THE THIRD GROUP WHICH
COULDN'T BE PAIRED
UP.
Figure 22a
different groups. When this highest power of two has been

thus paired, the successively smaller powers of two must
be similarly paired. (No power of two should be considered
until all larger powers have been properly treated.) But
the pairs of smaller powers of two are not required to be
in the same groups as the larger powers of two.
Eventually, in an unsafe position, a power of two
will be found which cannot be paired. Any of the groups
which contain such a pOWf'r of two can be chosen as the
Symmetry
77
group in which to operate. The remaining powers of two

are paired, making no use, so far as is possible, of the
chosen group. At the end of this process, the counters
remaining in the chosen group are removed. This move
will be the same as the one determined by the previously
described method. Figure 22a shows an unsafe position, and
the pairings that can be mentally chosen.
The binary representation for each of the groups of
matches in the original position, and the decimal addition,
are shown in figure 22b, and the representation for the safe
1001
1000
100
11
2112
1001
1000
10
11
2022
Figure 22b
position is also shown. The second method for winning

at Nim emphasizes the use of symmetry in determining
the proper move.
Strictly speaking, the game described above should

be called Niml because there exists a generalized version
called Nimk. Just as in Niml we are allowed to take as
many counters as we wish from anyone group, so in Nimk
we may take from as many as k groups if we wish. To
win at Nimk we write the binary numbers as before, but
now a safe position is defined as one in which each number in the decimal sum is divisible by (k + 1). (Niml is
seen to be a special case of Nimk.) The way to choose the
group or groups to operate on is a generalization of the
previously described method, and the limits at which this
method must be modified are apparent.
78
Symmetry
The 21 Counters
Another game, similar to Nim, is the game in which

a pile of 21 (or more or less) counters are supplied. Two
people play the game and each person, at his turn, takes
from the pile any number of counters less than a preassigned number, say 5. The object is either to take the
last one or make the opponent take the last one. This
game is very easily analyzed. The "key number" is 6. If
the opponent takes c counters, then we can always take
6 - c, regardless of what c is. In this way, control can be
maintained. If we wish to take the last counter, then we
must leave, for the opponent, some multiple of 6. If we
wish the opponent to take the last counter, we must leave
one more than a multiple of 6.
Grundy's Game
Grundy's Game is another game which is played with

a pile of counters. Each person, in turn, must choose a
group, and divide it into two unequal parts. The first
move divides the entire pile of counters into two unequal
parts. He who can no longer make a legal move loses.
If the game is played with a pile of 21 counters, the
proper strategy to be followed in order to win at this
game can be divided into two parts. First, we see that if
we leave a pile containing 4, 7, or 10 counters to our
opponent, then we can win. (When a single counter or a
single pair of counters is left, it can be considered as
"dead," because no further moves can change it. Thus,
a group of 3 can be divided into 1 and 2, but then a further
move is impossible.) Second, if two equal groups of counters are left to our opponent, then we can win, because
whatever our opponent does in one group, we can do in
the other.
Symmetry
79
A good strategy is to leave to the opponent a grollp

of 4 or 7 or 10 counters, or any groups of these numi)('rs,
or any two equal groups which may be in combination
with any or all of the above numbers.
Further investigation shows that (3, 6) counters, or
(4,7), or (5,8) are also safe numbers of counters to leav('
to the opponent. (This can be shown by a process of exhaustion.) Since this is so, the first player can always win
(we started with 21 counters). The first move should be
to leave (9, 12). Then whatever the opponent does in any
one group we can do in the other. In this way, the two
numbers nand n + 3 can be reduced by taking equal
quantities from each group, and this process must ultimately result in equal piles of counters, plus either (3, 6)
or (4, 7) or (5, 8). In any of these cases, the game can
easily be won. The analysis for Grundy's Game for an
arbitrary starting position can be very complex indeed. It
has been analyzed by T. H. O'Beirne in the British publication The New Scientist, and O'Beirne discusses this
strategy in a book about to be published by the Dover
Publishing Co. It does seem a pity that such a beautifully
simple game has such a complex strategy.
Mutilated Chess Board
A fairly well-known, and almost standard, "war horse"

type of problem is that of the "mutilated chess board,"
as Martin Gardner calls it. (It is not our intention to
compile many standard problems which have been described elsewhere. But the new problem bears such a likeness to the mutilated chess board problem, that it is only
fair and instructive to mention the previous problem.)
This problem concerns a board divided into 64 squares.
The two squares at the opposite ends of a main diagonal
arc cut out. Can the remaining area be completely covered
80
Symmetry
with 31 dominoes, each of which is two squares long and

one square wide? The answer to this problem is "No."
The proof is very beautiful, and almost devastating in
its simplicity. Since the squares are alternately colored,
each domino must cover a black and a white square. The
number of covered white squares must be equal to the
number of covered black squares. But the squares at opposite ends of a main diagonal are of the same color, so we
are left with 32 squares of one color, and 30 of the other
to cover. This cannot be done with 31 dominoes.
Construction of a Cube
The new problem involves the assembly of twentyseven 1 X 2 X 4 bricks into a 6 X 6 X 6 cube. This problem was told to me by Gary Gordon of RCA-AE Division,
and was originally invented by Professor R. Milburn of
Tufts University. The question is simply: Can the 6-unit
cube be formed with these 27 bricks?
(6 X 6 X 6
216
27 X 8)
To see that such a construction is impossible (even

though the combined volume of the 27 bricks is the same
as that of the desired cube), we consider the 6 X 6 X 6
cube to be colored as is shown in figure 23.
The individual cubes are colored in 27 groups of 8
each, and each group of eight is in the form of a small
cube. (It may be convenient to consider the full cube to
be subdivided into 27 2 X 2 X 2 cubes rather than one
6 X 6 X 6, and for it to be colored as a three-dimensional
checkerboard would be.) Each brick must necessarily
occupy an equal number of the small black and white
cubes. (This can best be seen by trying several different
placements.) These bricks can surely not form portions of
the 6 X 6 X 6 cube, which does not contain equal numbers
Symmetry
81
Figure 23
of small black and white cubes. But the full 6 X 6 X 6

cube contains 14 X 8 cubes of one color, and 13 X 8 of
the other. Hence, it cannot be formed of the 1 by 2 by
4 bricks.
Q.E.D.
9: Parlor Tricks and Number Manipulation
Eight 8's
Can you take eight 8's and arrange them in any

fashion connected by the usual signs of mathematics so
that they represent exactly one thousand? This question
appeared recently in a popular magazine. The big question
appears to be what, exactly, do we mean by "the usual
signs of mathematics"? Do we mean only the plus, minus,
times, and divide signs? Some people allow the use of the
square root sign a finite number of times j others would
allow its use an infinite number of times. We can consider
the use of the factorial function and the gamma function.
The implication in the problem, as I originally saw it, was
that there was only one solution. This is very false. (If
ever a superlative adjective needed to be compared, this
is it.) The solution they were thinking of was the following:
888+88+8+8+8= 1000
A much simpler solution would have been:
8888 - 888 = M (1000)
8
Other solutions, which are only slightly more complex, are:
8+8
- - (8 X 8 X 8 - 8) - 8 = M
8
82
Parlor Tricks and Number Manipulation
8(8 X 8 + 8 X 8) - 8 - 8 - 8
[(8 + 8)8 -
+ : + 8)J 8
83
Still other solutions (which are equivalent to 103) are:
88 - 8)(8+8+ 8 )/8
( 8 + 8)(8+8+ 8 )/8
(=Mand8+-=M
If we use the factorial sign, we can get:
_ 8 = M
8!
8(8 _ 8 +: + 8)
8'
8' - 8[(8 X 8 X 8) - 8] - 8 = M
8! [8(:8+_88)J - 8
If we now allow ourselves to use the gamma function, we
can get:
-~) -
8r(8)
_ 8
8+8+8+8+8
r(8) - (8 X 8)(8.8
We can use logs (base 8) and get:
[logs 8 X 8][8 X 8 X 8 - 8] - 8
[log8x8x8x8x88]r(8) - 8
[(88; 8)JOgs (8X8X8)] = M

8 + 8)J[IOgS (8X8X8)]
[ 8+ ( -8=M
84
And finally, we use the combinatorial symbol, where
(n) ==
T
I(
T.
n~
) and get:
I
T.
[ef) -8]
8~ M
+8 -
Possibly the most outrageous array of all is:

( 88 X 8 X 8 X 8)
88
8
1000
Of course, this last isn't really legal because it represents

one-zero-zero-zero in the number base 8. It looks like a
thousand, but looks are often deceiving.
The Multiplication Theorem
The Multiplication Theorem of algebra, which is

learned by all high school students (and forgotten by
most), is very important for an understanding of many
simple mental calculations. In one of its forms, it states
that the product of two factors (a + b)(e + d) is simply:
(a
+ b)(e + d)
= ac
+ ad + be + bd
(1)
The case where a = e and b = d is a special case which

deserves some note. It represents (a + b) (a + b) or
(a + b)2. We can write:
(a
+ b)2 = a 2 + 2ab + b2
(2)
The case where a = e and b = -d is another important

and special case. This case reduces to:
(a
+ b)(a -
b) = a 2
b2
(3)
Application of these three equations can be very useful in

many examples of mental calculations.
85
As a demonstration of the usefulness of equation 3,

the product of two numbers which differ by small, but
equal, amounts from a number whose square we know,
can quickly be stated. Thus:
19 X 21 = (20 - 1)(20 + 1) = 400 - 1 = 399
It is hoped that most people know 20 2 by heart. This
little trick does not seem to be of much interest in itself,
but it becomes more useful when one realizes that not
only are numbers ending in a zero easy to square, but
numbers ending in a 5 are also easy to square. To square
any such number, say 75, we merely multiply the number
without the 5-(i.e., 7 in this case) by the next higher
number (8), and then affix 25 to the end of the resulting
product. Thus:
7X 8
56
75 2 = 5,625
An example which utilizes both equation 3 and this easy

method of squaring numbers ending in 5 is the following
We can write:
73 X 77 = (75 - 2)(75 + 2)
= 75 2
4 = 5,625 - 4 = 5,621
The generalization of this method of squaring numbers

ending in a 5, to the multiplication of any two numbers
ending in a 5, is reasonably simple and quite useful. In
this latter case, we multiply together the two numbers
without the 5's, and add to this product half the sum of
the same two numbers. Then we affix 25 to the end as
before. Thus:
35 X 55 =?
3+5
3X5=15; -2-=4;
15+4=19;
35X55=1,925
If the sum of the two numbers without the 5's happens
to be an odd number, we take account of the
in the sum
86
by affixing 75 to the end of the number rather than 25.

For example:
35 X 65 = ?
3+6
3 X 6 = 18'' 2- = 4.!
2,
18
+ 4! =
22!i 35 X 65
2,275
The proof of this method of multiplying numbers ending

in 5 is simple.
Let us represent one of the numbers ending in 5 by
(lOa + 5), and the other one by (lOb + 5). Using equation (1), the product of these two numbers is:
(lOa
=
+ 5) (lOb + 5)
lOOab
= 100ab
+ 50a + 50b + 25
100
+ """"2
(a + b) + 25 =
(
100 ab
+-b) + 25
+ a-2
The particular case a = b, i.e., the case of squaring a

number ending in 5, gives the result:
lOO(a 2
+ a) + 25 =
100a(a
+ 1) + 25
These results agree with the aforementioned rules.

These little tricks were particularly useful to me one
day when a group of friends and I had worked out a
problem in probability. The problem involved an "electronic decimal scaler," which consists of 5 columns of the
digits from 0 to 9. The question was, if a perfectly random
number appears in each of the 5 columns, what are the
chances that it will represent nothing in poker, not a pair,
not 3 of a kind, 4 of a kind, 5 of a kind. The 5 numbers
could represent each of 5 cards in a game of poker with
no suit designations. To calculate the probability of getting
nothing we must recognize that the first number can be
anything, the 2nd number cannot be the same, so it must
be any of the remaining 9. The 3rd number must be one
87
of the remaining 8. The overall probability of getting

nothing will be (9 X 8 X 7 X 6)/104. This product, I stated
immediately, was .3024. I told my stunned and impressed
friends that I happened to notice that 7 X 8 = 56 and
6 X 9 = 54. The answer was obviously 10--4(55 2 - 1) =
.3024.
Some Useful Facts
Some other interesting facts concerning "smallish"
numbers (i.e., <50) and their squares, products, etc. are:
If one knows the square of a particular number (n), then
the square of the next higher is simply 2n + 1 greater.
In this way:
112 = 10 2 + 20
+ 1 = 121
And the next higher square must be 4n

122 = 102 + 40
+4 =
+ 4 greater, e.g.:
144
Another occasionally useful fact is that the square of a

number (25 - n) differs from the square of (25 + n) by
an integral number of hundreds, equal to lOOn: e.g.,
242 = 576, 26 2 = 676, or 23 2 = 529, 27 2 = 729. This fact
can easily be shown using the Multiplication Theorem:
+ n)2 = 25 2 + 50n + n 2
(25 - n)2 = 25 2 - 50n + n 2
(25
Difference
lOOn
In this manner, the squares from 26 to 49 can be simply

gotten from a knowledge of the squares from 1 to 24.
A trivial but real example of how these methods can
give startling results is the following true story. I once
jokingly asked a fellow student, while we were studying
in the same room, if he knew how much 2 20 was.
"Sure," he said, "it's 1,048,576."
88
The fact that I was impressed was obvious. He got

the answer so quickly by several rapid applications of the
Multiplication Theorem. He knew, as do most mathematically inclined people, that 210 = 1,024. Hence,
2 20 = 1,024 2 = (1,000
=
+ 24)2
1,000 2 + (2 X 24 X 1,000)
= 10 6 + (48 X 103)
+ (24)2
+ (24)2
242 is easy to obtain by one of our previous methods.

Hence:
2 20 = 1,024 2 = 10 6 + (48 X 103) + 576 = 1,048,576
Several years later, my friend confessed that, upon
realizing his success in this case, he memorized the population of Alaska. He hoped to secure his reputation of
being a wonderfully wise person, just in case I happened
to wonder aloud what the population of Alaska was!
Many so-called "magic" tricks of a mathematical
twist are easily proved with the aid of the Multiplication
Theorem. One well-known trick, for example, requires a
person to choose any three-digit number in which the
first and third digits are not the same. He is then asked
to reverse the three-digit number and fi~d the difference
between the reversed and original numbers. He is then
requested to reverse the digits of the difference, treating
it as a three-digit number if it is less than 100, and add
the resulting number to the original difference. The unsuspecting accomplice now has the number 1,089, regardless of the original three-digit number. He could, for
example, be requested to multiply the 1,089 by 40, because
that might be the number of wives that a sultan, who has
just consulted his prime minister concerning his insomnia,
had. The mathematical fact, 1,089, can obviously be
clothed in a great many stories. To prove that the final
sum is 1,089, we can proceed as follows:
89
The original number, abc, can be represented as:

10 2a
+ lOb + c
or
looa
+ lOb + c
Reversing this number yields:

10 2c
+ lOb + a
or
l00c
+ lOb + a
And the difference is:

100(a - c)
+ (c -
a)
We can consider only a > c. Remembering that (c - a)

is then negative, we modify the expression for the difference so as to utilize only positive numbers.
We modify this by subtracting 100 and then adding
90 and 10. The difference can thus be written:
100(a - c)
+ (c -
a)
lOO(a - c - 1)
+ 90 + (10 + c -
a)
(1)
This difference can now be reversed:

(10
+c -
a) is the first digit
(a - c - 1) is the last digit
The reversed difference is:

100(10
+c -
a)
+ 90 + (a -
c - 1)
(2)
The sum of (1) and (2) is:
+ 10 + + (10 + , - " + ~ - l00(~ - ; - 1
= (9 X 100)
+ 180 + 9 =
~)
+ (2 X 90)
1)
1,089
The multiplication theorem explains many simple socalled "magic tricks." It does not work, however, on a
trick which I once played on my younger daughter. I
asked her to think of a digit from 1 to 9. I then asked
her to open her mouth, and after examining the inside of
90
her mouth carefully, I triumphantly announced that the

number she had chosen was 6. In shocked disbelief she
asked, "But Daddy, how did you do it?" She was so
impressed that I didn't have the heart to tell her how
lucky my guess was. Incidentally, there is a 10 percent
chance of succeeding in this trick, and one's standing is so
increased if he guesses correctly, that it seems worth while
to try this scheme. (Keeping one's fingers crossed might
help a little.)
Casting Out Nines
Another example which the Multiplication Theorem

explains simply is the method of Casting Out Kines.
Once familiar to all school children, this method is a
simple check on the possible correctness of long multiplication problems. The sum of the digits in the multiplicand
and the sum of those in the multiplier are first determined.
If either of these sums has two or more digits, we take the
sum again and again until there is a single digit left for
each. This single digit is known as the digital root, or
simply the digital, of the original number. The product
of these two digitals must be equal to the digital of the
product. If it is not, the multiplication must be wrong.
If they are equal, the multiplication could be correct, but
is not necessarily so. To illustrate the method:
1234
X 5678
9872
8638
7404
6170
7006652
(digital) 1
(digital) 5
(product)
(digital) 7
+2+3+4 =
+6 +7 +8 =
10 -+ 1
26-+8
lX8=8
+0 +0+6+6+5 +2 =
26-+8
Therefore this product 7,006,652 might be correct.
91
The proof of this method is straightforward. To start,

we recognize that any number at all can surely be written
as the sum of some integral number of nines plus some
remainder. (It could also be written as some integral
number of eights or sevens, plus some remainder, but let
us consider only the nines.) We can thus write the multiplier Nl and the multiplicand N2 as:
The product NIN2 is, by equation 1:

NIN2
81nl~
+ 9n R2 + 9~Rl + RIR2
1
The most interesting feature of this expression is that each

term in the product, except for the product R 1R2, is divisible by nine. This is the same as saying that the remainder,
after dividing the product by nine, must be the same as
the product of the remainders of the original numbers
when they were divided by nine. But the remainder of
any number after dividing by nine is simply the digital
of that number. Because of this eagy way of getting the
remainders after division by nine, this method of checking
multiplication is quick and easy.
Other parlor tricks may require more or less ingenuity
to explain than the illustrated examples do, but most can
be explained in terms of the Multiplication Theorem.
10: Divisibility by Various Numbers
Divisibility by 2, 5, and 10
Divisibility of numbers by certain other numbers can

often be checked mentally-very quickly. For example,
everybody knows that if a number ends in a zero, it is
divisible by 10. And likewise, if a number ends in a zero
or a 5, it is divisible by 5. The reason is simply that any
number multiplied by 10 ends in a zero, and any number
multiplied by 5 ends in either zero or 5. If a number is
divisible by 2, it must be even.
Divisibility by 3
Only if the sum of the digits comprising a number is

divisible by three is the original number also divisible by
three. The reason for this is most easily seen by considering
what is really meant when we write a multidigit number.
If the number is represented by abcde (where a, b, c, d, e
are separate digits), we recognize that this is shorthand
notation for:
a10 4
+ blO + cl0 + d10 + e'lO = abcde

2
Any power of 10 has a remainder of one when it is divided

by 3. Hence, any of the terms in the above expression,
e.g., b .10 3, will have as a remainder when divided by 3 its
92
Divisibility by Various Numbers
93
coefficient (here b). Since the remainder after dividing

each term by 3 is simply the coefficient, the remainder
after dividing the whole number by three is simply the
sum of the coefficients. If this sum is divisible by 3, then
the original number must be divisible by 3.
Divisibility by 9
A method for checking and proving divisibility by 9

is also well known. Since any power of 10 has a remainder
of one when divided by 9, it follows that the digital of a
number must be divisible by 9 if the original number is
divisible by 9.
Divisibility by 11
A rule for divisibility by 11 is not quite so well known

as the two previous rules. In deriving this rule, we note
that any even power of 10 (e.g., 10, 10 2, or 104) has a
remainder of one when divided by 11. And any odd power
of 10 (e.g., 101 or 10 3, etc.) has a remainder of 10 when
divided by 11. Since a remainder of 10 can be considered
to be a remainder of -1 (as far as divisibility by 11 is
concerned), we conclude that the remainder left when
dividing abcde by 11 will be:
a-b+c-d+e
Simply stated, the remainder is the difference between the
sum of the digits in the odd places, starting from the right,
diminished by the sum of the digits in the even places. For
example, if 12,396 is divided by 11, the remainder is
(6 + 3 + 1) - (9 + 2) = 10 - 11 = -1, or 10, which
is the same thing. Only if the remainder is zero or divisible
by 11 is the original number evenly divisible by 11.
94
Divisibility by 7
Rules for divisibility by other numbers can be derived,

but they are unfortunately nowhere near as neat and
simple as the rules for 10, 5, 2, 3, 9, and 11. As an example
of one of these rules, let us consider the derivation of a rule
for divisibility by 7. Let the original number be abcdef. As
before, we write this out in full as:
(a X 105)
+ (b X
10 4)
+ (c X
+ (d X
103)
10 2)
+ (e X
101)
+ (f X
1(0)
The remainder after dividing 10 by 7 is 1, and so the

remainder after dividing (f X 100) by 7 is (f X 1). In the
same way, the remainder after dividing 101 by 7 is 3, and
that after dividing (e X 101) by 7 is 3e. In this way, we
can calculate the remainders for each term in the above
number, and the remainder of:
(a X 105)
+ (b X
10 4)
+ (c X
+ (d X
103)
10
2)
+ (e X
101)
+ (f X
10)
is:
(a X 5)
+ (b
X 4)
+ (c X 6)
+ (d X
2)
+ (e X
3)
+ (f X
1)
(The remainders repeat periodically as we go to the left

after a5, i.e., the next number is multiplied by 1, the next
one by 3, etc.) Thus, in the previous example of 12,396, the
remainder after division by 7 will be:
(1 X 4)
+ (2 X
6)
+ (3 X
2)
+ (9 X
3)
+ (6 X
1) = 55
And the remainder of this is 6 as can be seen in many

ways. For instance, (5 X 3) + (5 X 1) = 20 and the
remainder of 20 is (2 X 3) + (0 X 1) = 6. We can also
see it by simply noting that 55 = (7 X 7) + 6.
The above formula can be made slightly (but only
slightly) more presentable by replacing some large positive remainders by smaller negative ones. Since a remainder
95
of 6 is the same as a remainder of -1 (as far as divisibility

by 7 is concerned) and a remainder of 5 is the same as -2,
and 4 the same as - 3, we can rewrite the remainder as:
- 2a - 3b - c + 2d + 3e + f = the remainder of abcdef
after division by 7. As before, the remainder of 12,396 is:
+ (2 X 3) + (3 X 9) + (1 X 6) = 34,
and the remainder of this is (3 X 3) + (4 X 1) = 13 or 6.
- (3 X 1) - (1 X 2)
By the simplified formula, as with the original formula, the

remainder after division of 12,396 by 7 is 6. The test for
divisibility by 7 can be put into somewhat more concise
form, by the following scheme: Divide the total number
into groups of three digits each, starting from the right.
In each one of these groups, multiply the right hand number by 1, the middle number by 3, and the left hand number by 2. Then, the sum of the alternate triads diminished
by this sum of the intermediate triads is the remainder.
For example, in the number 12,396, we first write:
o1 2
3 9 6
"--"" "--""
We then mUltiply by 2, 3, and 1 to get:
o3
2 6 27 6
"--"" "--""
The alternate sum in this case is simply the difference of
6 + 27 + 6 and 0 + 3 + 2 ~ 39 - 5 = 34.
The remainder of 34 is 0 9 4 ~ 13 ~ 0 3 3 ~ 6
"--""
"--""
Most people will agree that this is not a cute and elegant
method. But it works.
Any number of the form abcabc, where a, b, and care
any integers from 0 to 9, is divisible by 13. This statement
is sure to be greeted with skepticism by some people. Not
only, however, is it divisible by 13, but it is also divisible
by 7 and 11. It is most convenient to write out the complete expression for which abcabc is the abbreviation. The
complete expression is:
96
abcabc = a 105 + b 10 4
In this expression, we
finally arrive at:
abcabc
+ c 10 + a 10 + b 10 + c' 10
can factor out (10 + 1) and we
l00l(a10 2
+ b10 + c100)
1
The number 1001 has the prime factors 7,11, and 13. Since
1001 is a factor of the number, so are 7 and 11 and 13 and
any multiplicative combinations of these numbers. This
fact can be used in a great many different ways to form
interesting problems. We could, for example, ask a person
to write down a 3-digit number, and then to repeat those
same three digits to form a 6-digit number. Then, we look
at it and we can say, "Aha! This number is divisible by
143." (We know that 143 is the product of 11 and 13, but
we don't have to tell it to the victim.)
We could also look at the number and say, "Oh, that's
divisible by 77." The fact that 7 and 11 are known factors
and that the product 77 is, therefore, also a factor doesn't
have to be stressed. This idea is used by Martin Gardner
in his "Mathematical Games" column in SciP:ntific American magazine to describe an amusing parlor trick.
The "wizard" asks someone to write any 3-digit number on a piece of paper. Then, a second person is asked to
repeat the same 3 digits to form a 6-digit number, and to
divide that 6-digit number by 7. [We know that the remainder will be 0, but the mystery may be introduced by
noting the remainder (0) on a piece of paper and giving
it beforehand to another person to hold.] A third person is
asked to divide the second person's answer by 13, and to
note the remainder of that operation. And then, the first
person is asked to divide that second quotient by 11 and to
note that remainder in the same place. If this first person
reacts as most people do, he will be shocked to find that
the quotient he obtained is the 3-digit number he first
wrote down. The other two people will be surprised to a
97
somewhat lesser degree when they discover that the result

of both of their divisions had a zero remainder.
Another similar problem is the following: Prove that
if we take any number and rearrange its digits to form any
other number, the difference between the original number
and the second number is always divisible by 9. We shall
consider the first arbitrary number to be represented by
abcdefg. The rearranged number might be fadbecg. The
difference between these two numbers must necessarily be
a sum of terms, each of which consists of one of the original
digits times the difference between two powers of 10, e.g.,
a(10 6 - 105) or b(1OS - 103). The difference between two
powers of 10 is always either zero or a string of nines followed by a string of zeros. Either case is divisible by 9 and
so the whole number must be divisible by 9.
As our last example, we can consider the following
problem:
Examine the number:
8, -5 - ,383, -2 - ,936, -5 -,8 -2,03- ,9-3, -76
Each of the ten blank spaces in the above number is to be

filled in with a different one of the digits from zero to nine.
What is the probability that the reSUlting number will be
divisible by 396?
The first reaction to this problem is that it represents
the height of idiocy or the acme of stupidity, whichever
you prefer. Actually, however, the problem solver has the
advantage in that we suspect that no one in his right mind
would even propose such a problem unless it had a reasonably clever solution. This is actually the case. First, we
notice that the number ends in a 76. So the total number
must be divisible by 4. Secondly, the sum of the digits
from zero to nine is known to be 45, and the sum of the
digits that are given is equal to 90, so that the sum of all
the digits is 90 plus 45 = 135, regardless of the order in
98
which the digits from zero to nine are inserted. Since 135
is divisible by 9, the original number must be divisible
by 9. It will be noticed that the blank spaces occur with
Machiavellian regularity only in the second, fourth, sixth
places, etc. - always in the even places. We can thus determine that the sum of the digits in the odd places is 73, and
the sum of the digits in the even places is 17 plus the sum
from zero to nine, or 17 plus 45 = 62. Since the difference
between 62 and 73 is 11, the original number must be
divisible by 11, regardless of the order in which the digits
from zero to nine were inserted. The known factors are
thus 4,9, and 11, and the number is therefore divisible by
the product of these numbers, 396. The probability is thus
seen to be unity.
11: Problems Which Seem to Have

No Solution
The Careless Indian Who Drops His Paddle
An Indian, canoeing upstream, accidentally drops

one of his paddles. He does not realize this until ten minutes
later. At that time, he turns around and paddles downstream (at the same rate relative to the stream), and
catches up with the paddle one mile from the place he
dropped it. What is the rate of the stream?
One of the easiest ways to solve this puzzle is to consider the flowing stream to be a conveyor belt, and the
Indian to be walking on it. On the belt, he drops the
paddle, walks away for ten minutes, and then returns. His
total trip has taken him twenty minutes, and the paddle (or
the conveyor belt) has moved one mile in this time. The
rate of the stream is one mile in twenty minutes or three
miles per hour.
The Early Commuter
A man commutes to work, and each day his chauffeur

leaves the house in time to meet his master at the station at
exactly 5 o'clock. The chauffeur then turns around (instantaneously, of course) and takes the man home. One
day, the man comes into the station early, and starts to
99
100
Problems Which Seem to Have No Solution
walk home. On the way, he meets the chauffeur, who had

left home at the usual time. He gets in the car and arrives
home ten minutes earlier than usual. At what time did he
meet the chauffeur?
The chauffeur drives at the same speed going to and
from the station. If the chauffeur got home 10 minutes
early, then he must have saved five minutes going to and
coming from the station. Since he normally arrived at the
station at 5 o'clock, he met his master and turned around
at five minutes before 5.
This solution can be represented graphically as is
shown in figure 24. In this figure, the path AB represents
the chauffeur's normal trip to the station, and BC the
trip back. The time at B is exactly 5 o'clock. On the day
in question, the chauffeur travels AED. DC is given as
ten minutes. Since ABC and AED are both isosceles, E
must occur at five minutes before 5 o'clock.
STATION
DISTANCE
HOME
TIME __
Figure 24
The Officer and His Column of Men
A column of soldiers is one mile long. The officer in
charge, standing at the rear of the column, gives the order
to advance. He starts to march in the same direction, but
at a faster speed. When he reaches the head of the column,
101
he turns around and marches toward the rear. When he

reaches the tail end of the column, he notices that the tail
is now where the head was when they started. How far
has the officer marched?
The diagram in figure 25 shows the officer and the
column of troops at three different times.
COLUMN
e..-
t=o
OFFICER
COLUMN
t =t,
OFFICER_e
COLUMN
t=t2
.... OFFICER
Figure 25
Let us denote the velocity of the column by Vc and

the velocity of the officer by Vo. The rate of the officer with
respect to the column when they are going in the same
direction is (vo - vc). So tl must be the length of the column
(one mile) divided by this velocity, i.e., l/(vo - vc). The
rate of the officer relative to the column when they go in
opposite directions is (vo + vc). The time interval (~ - t1)
is l/(vo + vc). The total time from zero to ~ is then:
I
~=--+-Vo -
Vc
Vo
+ Vc
Time from 0 to t2 is also equal to l/ve because the column

advanced one mile. We can then write:
102
=Vo - +Vo-+-Ve
Ve
- Ve
Let us multiply by
Vo
and let d
= volve:
Vo
Vo
d=--+-Vo -
Ve
Vo
+ Ve
Dividing numerator and denominator by
Ve
we get:
d=d-1+d+1
We recognize that d
officer walked:
(d 2
= [Vo
X (live)] is the distance that the
+ d + d2 -
Simplifying, we get d 2
d)1(d 2
1).
2d - 1 = 0
VS) 12 = 1 V2
The officer walked 1 + V2 miles.
d = (2
The Master Painter Ve. the Apprentice*
A master painter can put two coats of paint on a particular barn in one day. He normally paints one coat before
lunch and one coat after lunch. Each coat takes the same
length of time to apply. One day he hires an apprentice
two hours before lunch. He fires him after the first coat is
entirely applied. He then goes on to finish the barn himself,
observing the usual lunch time (he uses fast drying paint,
so that he can start the second coat immediately). One
hour before his usual quitting time, he notices that he has
applied the second coat everywhere except where the apprentice had painted in the morning. The apprentice
* This problem, incidently, was invented by V. D. Landon to

fill an apparent void in such problems coming to our attention.
103
paints 200 square feet per day. What is the area of this
barn?
The area that the apprentice painted in the morning
must have been as much as the master painter could have
painted twice (two coats) in one hour. The master painter
would then have taken one half hour to paint this area
once. The master painter would have finished one coat on
the entire barn himself by lunch time. So he would have
painted everything except what the apprentice did by half
an hour before lunch. (The apprentice must, therefore,
have been fired half an hour before lunch.) In other words,
the apprentice took I! hours to paint what the master
painter could have done in ! hour. Since the apprentice
paints 200 square feet per day, the boss can paint 600
square feet per day j the area of the barn must be 300
square feet.
The Two Ferry Boats
There are two ferry boats, Fl and F 2, on opposite sides

of a river. They travel at different rates, Vl and V2. They
start to cross the river at the same time and pass each
other 720 yards from one side. Upon reaching the opposite
--1=0
ITIME
BANK I
BANK 2
I~'------W------~'I
Figure 26
-1=12
104
banks (at different times, of course) they turn around and

go toward the opposite shore. This time they pass each
other 400 yards from the other bank of the river. How
wide is the river?
The diagram on figure 26 shows how equations can
be simply derived.
Since the boats start at the same time, we can write:
or:
V2
Vl
720
w - 720
(1)
We can also write for the second intersection at

v2~=w+400
or:
V2
Vl
Vl~ =
~:
2w - 400
w + 400
2w - 400
(2)
Equating (1) and (2):

720
w - 720
w + 400
2w - 400
1440w - (720 X 400) = (w
+ 400)(w -
1440w - (720 X 400) = w 2
720)
320w - (720 X 400)
The large constant terms cancel, and we are left with:
w = 1760 yards
Volume of a Modified Bowling Ball
A standard size bowling ball has a hole drilled straight

through the center. The side of the finished hole is three
inches long. What is the volume of the bowling ball after
the hole has been drilled out? Protestations saying, "You
105
didn't give me the original diameter of the ball!" or, "You

didn't tell me the diameter of the hole (or the drill)" are
common. But these numbers are not required for a solution.
A straightforward calculation, using calculus, of the
volume of the ball remaining would show that this volume
depends only on the length of the side of the hole and not
on the diameter of the hole or the original diameter of the
ball. Another argument can be advanced by those who
prefer not to use the calculus. One can say, "This chap
doesn't give problems that have no solution. Since he
didn't give me the diameter of the hole, the volume of
ball remaining must be the same, no matter how big the
diameter is (as long as the side is three inches long). If
this is so, then I can assume that the hole is an exceedingly
thin hole (perhaps vanishingly small). And then the
volume of the ball remaining would be simply the volume
of a sphere whose diameter (the length of the vanishingly
small hole) was exactly three inches, and this volume is:
~ rRa = ~ r (~y = 92r
12: Probability Problems
Die Throwing
How many times, on the average, must one toss a sixsided die before every number from one to six comes up at
least once?
The solution to this problem is quite straightforward,
but perhaps it can be made even easier to accept if we consider a simpler version of the same problem first. Namely,
how many times, on the average, must a coin be flipped
before at least one head and one tail appear? We note
that the first throw must be either a head or a tail, and we
wish to know how many additional tosses are required, on
the average, before we get the other one. The probability
of getting the other one is t. If S represents the number of
successes and N the number of trials, probability can be
defined as the limiting value of the ratio SIN, as N tends
toward infinity. Or probability is SIN where S is the
average number of successes in N trials. In general,
p = SIN. In this problem, we want to know the average
number of tosses to get one success if the probability is t.
Since N = Sip, this number is II! = 2. Thus, the answer
to the simpler version is one (for the first toss) plus two (to
get the other one) equals three.
Returning to our original problem; on the first throw,
we'll get one of the six numbers, we don't care which one.
Having that one, the probability of getting a number
106
Probability Problems
107
different from the first one is t. On the average, then, t of a

throw are required to get something different from the first.
After this second number, the probability of getting something different from the first two is t, and we expect to
require -! of a throw to get it. In this way, the total number
of throws will be 1 (for the first) plus t (for the second)
plus -!, etc., or:
T (number of throws) = 1
+ t + -! + t + t + t
= 14.7
There are bound to be skeptics who do not believe this

solution. An experimental test of the coin tossing problem
is much easier than a test of the die throwing problem.
Bertrand's Box Paradox
In "Bertrand's Box Paradox" there are three boxes,

all of which look alike on the outside. One is known to contain two gold coins (gg) , the second, a gold and a silver
(gs), and the third, two silver coins (ss). One of these boxes
is chosen at random, and a single coin is withdrawn. That
single coin turns out to be silver. What is the probability
that the coin remaining in that box is also silver?
Most people reason in the following manner: If the
first coin turns out to be silver, it must have been taken
either from the ss box or the sg box. Since the probability
of choosing anyone box is the same as the probability of
choosing any other, and we know that one of only two
boxes must have been chosen, the probability that the
remaining coin is silver must be one half. This reasoning is,
of course, incorrect. (If the incorrect solution were not so
attractive, I would not have included such a simple example.)
The reasoning above is incorrect (even though many
famous mathematicians were fooled) because it is not
equally likely that the ss or the sg box has been chosen (if
108
we know that a silver coin has been chosen). It is, in fact,

twice as likely that the ss box has been chosen rather than
the sg. An easy way to see this is to consider the procedure
of picking a box and choosing a coin to be repeated a large
number of times-say, six thousand. Half of all these
times, a silver coin will be chosen, and each box will be
chosen a third of all the times. But in only half of the number of times in which the sg box was chosen will a silver
coin be withdrawn. So, we have two thousand cases where
the ss box was picked (and the silver coin withdrawn),
but only half of two thousand for the sg box. It is clear
from this analysis that in two thirds of the cases the ss
box has been picked.
Another method of reasoning: If the original statement had neglected to tell what kind of coin had been
picked, and had simply asked for the probability that the
remaining coin is the same as the first one, the answer
would clearly be two thirds, because we are interested in
either the ss or the gg box. If a golden coin were withdrawn,
the probability would be two thirds that the remaining coin
was gold. And similarly, if a silver coin were withdrawn,
the probability would be two thirds that the remaining
coin was silver. Since it makes no difference what kind of
coin was chosen, the statement that" the coin was silver"
contains no additional information and cannot change the
probability. The probability is thus two thirds.
Probability of Identical Pairs
Among Cards
Suppose we have two shuffled decks of cards face

down, and side by side, and we compare the first, second,
third, etc. cards of each deck. What is the probability
that at least one pair of these cards will be an identical
pair (e.g., queen of spades and queen of spades)? The
109
solution to this problem can be very messy and difficult

if we do not use the right approach. For the sake of simplicity, we can consider the cards to be numbered from 1
to 52, and only one deck need be shuffled-the other can
he in its natural order. The probability that the first card
bears the number 1 is -h (because there are 52 numbers
possible, and only one is correct). The probability that the
first card does not bear the number 1 is [1 - -hl. The
probability that each succeeding card does not correspond
to its order in the deck is nearly [1 - -hl. (It is only
"nearly [1 - -hl" because we have neglected the possibility that the number r has appeared before the rth card
was inspected. The probability would be unity (1) then,
that the rth card does not bear the number r. But the assumption that the probability is [1 - -hl so simplifies the
calculation of the total probability, that we yield to the
temptation.)
The probability that none of the 52 cards bears a
number corresponding to its order is the product of all
these probabilities, or [1 - -h1 52
The evaluation of this quantity is somewhat simpler
than it appears offhand. The expression [1 + (l/n)ln has
an unexpected behavior. It can be evaluated for n = 1,
2, 3, etc., and it is found that the value, instead of increasing beyond all bounds, approaches 2.71828 ...
This number is the celebrated number known as e.
If we consider the fraction [1 + (x/n)ln then this function
chokes off at a value of eO;. In the original expression that
we had to evaluate, namely, [1 - -h1 52, it can be shown
that the value is very close to e-1 = l/e. Since e is equal
to 2.71828, the probability that no identical pairs will be
found is somewhat greater than one third, and the probability that at least one identical pair will occur is somewhat less than two thirds. This rarely fails to shock people
unfamiliar with this problem.
no
Probability of Same Birthday

Among N People
How large a group of people must be considered in

order to have a 50-50 chance that some pair in the group
has the same birthday? In this problem, as in the last, it
is worth while to consider the probability that the event
will not occur. The first person can have any birthday
during the year. The probability that the second person
does not have the same birthday is ill. And the probability that the third does not have the same birthday as
either of the first two is lli. The probability that no pair
among a group of N people has the same birthday is:
364 363
362
365 - (N - 1)
P = 1 X 365 X 365 X 365 X ... X
365
If we set P equal to one half and solve for N, N turns out
to be nearly 23. So the chances are a little better than

50-50 that some pair of people in a group of 23 have the
same birthday. (Don't forget that any two of the group
might make up the pair, it might be the 6th and the 20th.)
St. Petersburg Paradox
The St. Petersburg Paradox is a game in which the

player tosses a coin, and keeps tossing so long as it comes
up heads. The game is over when he tosses his first tail.
The banker pays the player 2N dollars, where N is the
number of heads he tossed before the first tail. If the player
gets a tail on the first shot, he gets 2 = 1 dollar. The
question is: What is a fair price for the player to pay the
banker for each game?
To answer this question, we make use of a concept
known as the mathematical expectation. The expectation
is the sum of the products of the probability of each pos-
111
sible outcome times the gain for that event. In this case,
the probability of getting N successive heads and then a
tail is 1/2N +1, and the gain is 2N dollars.
Expectation
L probability X gain
Thus, the expectation is

0>
2N
0>
L 2N+l =N-O
L -21
N=O
1
= -2
+ -2 + -2 + ... + -2 + ...
and this expression clearly diverges. It is larger than any

finite number we can choose. The paradox is that the expectation should be the fair price to pay, but no one but a
fool would pay an infinite amount of money for the privilege of playing this game. What has gone wrong? Is our
calculation incorrect? In spite of the fact that the calculation is so simple, it does contain an error. It tacitly
assumes that if the player were to get 50 heads in a row,
the banker would be able to pay him 250 dollars. But this
is more money than exists in the world! Our calculation
must take into account the fact that there is a maximum
of money which the banker can pay. If we assume this
maximum is $1024 (for ten heads in a row) then the calculation of the mathematical expectation must be modified
in the following manner: The mathematical summation
must be broken into two parts. The first is calculated in
the usual way up until N = 10, and then we must add on
to this a summation of 1/2 N +1 times a fixed gain of 210.
This can be written:
Expectation
10
L -N X
N=o2 +l
2N
1
+N=1l2
L -N
+
0>
1
X 210 = 5 2
+ -21 =
and this says that the fair price for each game would be
six dollars. (This is quite a bit more reasonable than an
112
infinite sum of money, but it still sounds rather high.)

Another point we should recognize about this game is that
our expected earnings depend upon the number of times
the game is played. On the average, a string of five heads
and a tail would be expected only once in every 26 games,
and we would hardly be justified in expecting such a string
in four or five games. We can also calculate the average
number of throws per game. The expression for the average
number of throws per game is the sum of the probabilities
of the game lasting N tosses times N. So, on the average,
the number of throws would be equal to:
GXl) + GX2) + GX3) + C16X4) + ...

+C~ XN) +...
Count Buft'on was curious about the mathematical expectation in the St. Petersburg game. He arranged for a
child to play the game 2,048 times, and found that the
total payment (of bank to player) was 10,057 units. The
expectation is thus approximately 5. This is in fair agreement with the ideas expressed in this section.
EVALUATION OF
"'N
1 2
L-N
This expression can be evaluated by writing the sum

as shown on the facing page.
The vertical sum of the terms on the left is seen to be
equal to the series we wish to evaluate, and the sum on
the right, which is the average number of throws per
game, is equal to 2.
2 + 4 + 8 + 16 + 32 + ... + 2N + ...
1
8+ 16 + 32 + ... + 2N + ...
1
16
4 + 8+ 16 + 32 + ... + 2N + ...
1
+ 32 + ... + 2N + ... = 8
111
32 + ... + 2N + ... = 16
113
* (Read It to Find Out)
The number * has been associated with tragedy since time

immemorial. For this reason, the *th floor of a hotel is
often omitted. The actual damage brought on by lack of
respect for this number may be long delayed and very
long lasting-as, for example, in the case of the" Sleeping
Beauty." In this case, you remember, the wicked old
witch (who was the *th fairy, and whom the regents didn't
invite to their child's christening) cast a spell on the child
which didn't take effect until 16 years later, and which
lasted for one hundred years. (They don't make spells
like that any more.)
In view of the dangers associated with *, it would seem
prudent for authors to avoid such a chapter in their books.
But people have been blithely ignoring these dangers, and
have been writing *th chapters with no regard for the
awesome events of the past, and with no apparent ill
effects to themselves. It therefore seems probable that the
terrors exist mainly in folklore and superstition, and not
in reality.
But what's the point in taking chances?
* is intended to denote the decimal equivalent of the binary
number HOI.
114
14: Humor in Mathematics
Liars and TruthTellers

The problem of the liars and the truth-tellers is a
fairly old one. A missionary is shipwrecked on an island
where there are two tribes of natives-the liars and the
truth-tellers. The liars always lie when they speak, and
the other tribe always tells the truth. The missionary
wishes to go to the town of " A," but as he goes, he comes
to a fork in the road. He sees a native standing there, but
doesn't know to which tribe the native belongs. What one
question can the missionary ask to find out which road
to take, regardless of which tribe the native belongs to?*
The answer is that the missionary asks, "Which road
would another member of your tribe tell me to take if I
wanted to get to the town of 'A'?"
If the native were a truth-teller, he would tell the
truth. And, if he were a liar, he would lie about what a liar
would say, and thus he, too, would tell the truth. The
missionary finds out which road to take, but he doesn't
know whether the native was a liar or not.
* If you take me to task for ending this sentence in a preposition,

I will counter with a story about the small boy who was sick and in
bed. His mother brought a book from which she could read to him.
He didn't like the book, and he petulantly asked, "Mother, what
did you bring that book I didn't want to be read to out of up for?"
115
116
Humor in Mathematics
Liar or Truth-Tener-at N-tuple

Intersection
The same missionary on the same island comes to an
n-tuple intersection of roads. He sees the same native
(of unknown honesty). What one question can he ask the
native to find which road to take and whether the native
is a truth-teller or not?
The missionary asks, "Which of all these roads does
not lead to the town of' A'?" The liar will point to the one
road that does lead to" A," and the truth-teller will point
to all the other roads except the one that does go there. In
either case, the single road to take is clearly indicated, and
the multiplicity of the answer is inversely proportional to
the honesty of the native! One might think that Information Theory would forbid getting so much information
from a single question. But the fact that the question does
not have a simple yes or no answer makes the analysis
more complex than first meets the eye.
Where on Earth Was He?
A hunter travels 10 miles south, 10 miles east, 10 miles
north, and shoots a bear. He discovers that he is back
exactly where he started. What color was the bear? The
traditional answer is that the bear was white, because the
only place where one can walk as described and come back
to the starting point at the end is the North Pole. This
answer will undoubtedly satisfy many people, but the
question we ask is: how many such starting points are
there on the earth?
Consider a circle of latitude near the South Pole,
exactly 10 miles long. Then, all points on the circle of
latitude 10 miles north of this circle will be points that
fulfill the conditions of the problem. The hunter goes
117
south, around the lO-mile circle of latitude (ending where

he began) and then north to the point from which hpstarted. The upper circle of latitude, of course, contains an
infinite number of points. And if this infinity of answers
were not sufficiently deflating, we can consider those
circles of latitude 10 miles north of circles whose circumference is 10/2, 10/3, etc., miles. (But at any of these latter
points, his "bear" would have looked more like a penguin.)
Goldbach's Conjecture
A German physician named Goldbach noticed that

in every case he tried, an even number could be represented
as the sum of two, and only two, prime numbers in at least
one way. Thus:
20 = 7 + 13,
22 = 11
+ 11,
24 = 7 + 17, etc.
Goldbach wrote to the mathematician Euler mentioning

this curiosity, and asked if Euler could prove it. Euler is
said to have written back seventeen years later, saying,
"No." Although many famous mathematicians have tried
to prove the "Goldbach Conjecture," none of them has
had any success at all, until very recently.
The first partial success was reported by a Russian
mathematician named Schnirelmann, who proved, in 1931,
that any even number could be written as the sum of not
more than 300,000 primes! And to top this, another
Russian named Vinogradoff proved, in 1937, that nearly
every even number could be written as the sum of no more
than four primes. The meaning of the word "nearly" in a
mathematical proof could be that the fraction of all even
numbers less than N which could not be represented as the
sum of four primes approaches zero as N approaches infinity . Work on the Goldbach conjecture is still going on,
but there do not seem to be any startling new developments.
118
A "Seemingly Impossible" Problem
An article in the A merican Mathematical M onthly*

recently reviewed the methods used in Russia to teach
mathematics. At the end of the article there was a list of
problems typical of those given in Russian math finals.
One problem gave the sequence 100001, 10000100001,
1000010000100001, etc., and asked for a proof that all
numbers of this sequence are composite (i.e., not prime).
We can represent the numbers in this sequence by Gn
where n is the number of groups of four zeros in the number. It is easily shown that G1 is divisible by 11, G2 by 111,
G3 by 1,111. But G4 is not divisible by 11,111. Somewhat
more sophisticated methods can be used to show that G5
and all higher G's are composite. But G4 is adamant in
refusing to yield to a proof that it, too, is composite.
Many people tried their luck with this problem and had a
signal lack of success. But one day, John Waite, an engineer
at RCA, showed me two ridiculous numbers, and with a
grin broader than the one on the infamous cat, asked me
to multiply them. The numbers were:
4,672,725,574,038,601 and 21,401
To say that I was shocked to find the product equal to G4
ranks as an understatement comparable to General
Custer's reputed" Over that hill, fellows, I think there are
friendly Indians!"
The original problem may have been a "ringer," and
may have been included merely to see how far the youngsters could get. Or perhaps Russian educators didn't
realize how tough the problem really is. The method used
by John Waite in cracking this painfully large number is
still a mystery to me, but I know it leans heavily on
studies by Gauss and by Euler of the composition of very
large numbers.
* American Mathematical Monthly, 64 (1957),
p. 397, Problem 8.
119
Archimedes' Cattle Problem

"Archimedes' Cattle Problem" considers a herd of
eight differently colored bulls and cows, and gives seven
simple arithmetical relationships between them (e.g., the
number of white bulls equals [! + tJ the number of black
bulls plus the number of yellow bulls). It then asks for the
number of each of the differently colored oxen. The solution to this part of the problem is laborious, but within
reach, each number being of the order of several million.
The original problem, however, puts additional restrictions
on the numbers by saying the number of white bulls plus
the number of black bulls is a square number, and the
number of dappled bulls plus yellow bulls is a triangular
number. (A triangular number is one of the set 1, 3, 6,
N
10, 15 ...
L n, so called because they fit nicely into a

1
triangle, as for example, do ten bowling pins.) The number

representing the total number of cattle then contains
206,545 digits, as was shown in 1880 by A. Amthor.
Amthor remarked that a sphere of diameter equal to that of
the Milky Way could contain only a small part of this number of animals, even if each were as small as a bacterium.
Most people would consider the problem sufficiently
solved for their tastes (or distastes) at this point. But
A. H. Bell, a surveyor and civil engineer of Hillsboro,
Illinois, succeeded in convincing two of his friends to join
him in calculating the first thirty and the last twelve digits
of this gargantuan number. But the intermediate 206,503
digits are still unknown. I like to think there is a special
place in heaven reserved for people like Mr. Bell and his
friends, who work so hard and long (it was a four-year
job) for so little tangible reward.
Definition of a Catalyst
A catalyst is familiar to many as a substance which
must be present to enable a certain chemical reaction to
120
take place, but which does not take part in that reaction.
It is not always easy to give an example of a catalyst. I
have long been fond of the following illustration:
A wealthy Arab died, and his will stipulated that one
half of all his horses should go to his oldest son, one third
of his horses to his middle son, and one ninth to his youngest son. At the time of his death, the Arab owned seventeen
horses, and the sons sorely lamented the thought of carving
up some of these fine Arab steeds to comply with their
father's will. As they were bewailing their plight, an
itinerant Arab with a seedy old nag chanced by, and he
offered the following suggestion:
"Look, boys," he said, "let's put my horse in with
your father's seventeen. Then we have eighteen. You, son
number one, are to get one half of the horses; that is nine.
And you, son number two, are to get one third of the
eighteen-or six horses. And you, son number three, are to
get one ninth, or two. That's a grand total of seventeen,
and one is left over. And that's mine. So long, boysl"
The Arab's horse was a catalyst.
Mnemonics for the Value of Pi

The number Pi plays an important role in many
branches of mathematics. One of its definitions (and
probably the only one familiar to some people) is that it is
the ratio of the circumference to the diameter of a circle.
The value of 'II" has been calculated to extreme accuracy,
and mnemonics are often devised to help remember the
digits.
For example, in the sentence, "May I have a large
container of coffee?" the number of letters in each word
corresponds to the successive integers in the decimal
expansion of'll".
George Gamow once wrote an article for Scientific
121
American,* in which he mentioned the value of 7r, and gave

the first five decimals in the expansion as 3.14158. Several
months later, a reader wrote to Scientific American, chiding
Gamow for making a mistake in the fifth decimal place.
Gamow answered, apologizing for the error, and explained
that it was due to the fact that he was an atrocious speller.
I t seems that he remembered 7r from the number of letters
in each word of the French poem:
"Que j' aime Ii faire apprendre
Un nombre utile aux sages
Immortel Archimede artiste ingenieur
Qui de ton jugement peut priser Ia valeur
Pour moi ton probleme
Ales pareils avantages!"
He spelled the word "apprendre" with only one "p," and

this caused him to recall the value of 7r as 3.14158. He does
not guarantee the accuracy of this poem beyond the first
few lines.
A Short History of the Accuracy of Pi

The ratio of the circumference of a circle to its diameter (7r) has been known for a long time, and the history
of the accuracy to which it is known is fairly interesting.
In the Bible, the value of 3 is used. Archimedes knew
that 7r was less than 3;-, but greater than 3*. The value
some of us use today, 3.1416, was known at the time of
Ptolemy of A.D. 150. F. Vieta, in the sixteenth century,
calculated 7r to ten decimals, and at the end of the sixteenth
century, Ludolph van Ceulen calculated 35 decimals.
Van Ceulen requested, in his will, that these 35 figures be
engraved on his tombstone, and this was done. In Germany
the number is still sometimes referred to, in his honor, as
* Scientific American, Oct. 1955.
122
the Ludolphian number. More and more decimals were

calculated in the succeeding years, and Shanks, an English
mathematician, calculated 707 decimals in 1873. (Poor
fellow, he made a mistake in the 528th decimal place, but
never found out about it. All succeeding figures are, of
course, wrong.)
Shortly after this, Lindemann proved, in 1882, that
1(" was a transcendental number (not the root of a finite
algebraic equation with integral coefficients) and therefore
could not possibly be a repeating decimal. In spite of this,
the Eniac computing machine was used in 1949 to calculate
2,035 decimals of 1(", and E. T. Bell reports that "recently,
some enthusiastic idiots computed 1(" to ten thousand
decimals."* The existence of Lindemann's proof of the
irrationality of 1(" seems to make the reasons for calculating
so many decimals even less attractive than the reason for
climbing Everest. But there's no accounting for tastes.
Best of All Possible Mnemonics
One of the branches of physics treats the different

probability distributions. One of these distributions is the
Poisson distribution, which states that if m happenings
occur on the average, then the probability that n will occur
is simply Pm(n) = (mne-m)/n! where e has its usual value
of 2.7182 ... , and n! is n factorial. This expression is fairly
important (especially to a graduate student who knows
that his mean old professors will expect him to remember
it). It is customary, when seemingly arbitrary things have
to be learned, to try to devise mnemonics for them. The
* Recently the value of ". hll.'3 been computed to more than
100,000 decimals. (Shanks, David, and Wrench, John W., Jr.,
"Calculation of Pi to 100,000 Decimals," Mathematics of Computation,
Jan. 1962, Vol. 16, No. 77, pp. 76-99.) When, oh when, will this
madness end?
123
Poisson distribution has the best of all possible mnemonics*

-the first 6 letters of the word mnemonic itself:
(m to the n, e to the minus m, over n factorial)
Experimental Determination of Pi
A beautiful theorem, discovered by Euler, relates the

sum of the reciprocal powers of the integers to an unusual
product involving all prime numbers. The theorem states
that:
1
IT
all primes
ex>
_a
1- P
= L--;
1
where s is an arbitrary integer. L (lln a) is known as rCs).

In particular, r(2), that is, the sum of 11n2 = 1/' 2/6. It is
shown at the end of this chapter that 1/r(2) is equal to
the probability that two numbers chosen at random are
relatively prime. (This is only one of the vast number of
unusual and unpredictable relations existing in mathematics.)
R. Chartres did an experiment in 1904 in which he
asked 50 of his students to write down five pairs of numbers chosen at random. He found that 154 of the 250 pairs
of numbers had no common factor. From this, he could
deduce that 1/r(2) = 6/r2 = ill. This gives a value of 1/'
equal to 3.12. (It is quite apparent that this method of
calculation does not have high accuracy as one of its
attractions. )
At the Chicago World's Fair in 1893, there was a
mechanical implementation of Count Bufl'on's Needle
Problem. In this problem, a needle of length l is repeatedly
dropped onto a table on which are ruled parallel lines.
* This mnemonic was devised shortly after I had read, and

was impressed by, Voltaire's Candide.
124
These lines are separated by a distance L. The probability

that the needle, dropped at random, will intersect one of
the parallel lines is equal to:
twice the length of the needle
11" times the distance between the lines
2l
1I"L
There were two counters set up in Chicago, one to

count the total number of tosses of the needle, the other
to count the number of times a line was crossed by the
needle. It is said that, after a year of operation, the value
of 11" was approximately 3.16 .... This method is also not
recommended for its high accuracy. Much better results
were reported by Lazzerini. He tossed the needle 3,408
times, and got a value of 11" with an accuracy of 3 X 10-7
(3.1415929). This accuracy appears to be unreasonably
high, since, if there was a normal distribution, one would
expect the variation from the true value, divided by the
value itself, to be inversely proportional to the square root
of the number of trials. Since N, in this case, was 3,408,
one would expect the variation in 11" divided by 11" to be
1
about V3408 (between 1 and 2 percent).
There seems to be no doubt that Lazzerini was extremely fortunate in getting such high precision from so
few trials.
A Letter in the Physical Review
Some people may object that this book, which is not

written by a mathematician, pokes fun at mathematics.
Why doesn't the author poke fun at his own subject,
physics? To show that I am not maliciously biased, I can
tell a story about Dr. G. Gamow. He and Dr. R. A. Alpher
wrote an article for the Physical Review. Before submitting
the article, they persuaded Dr. Hans Bethe (pronounced
125
Bay-tuh) to join them as an author. The article, accordingly, appeared in the Physical Review letters, * entitled
"On the Abundance of the Elements" by Alpher, Bethe, and
Gamow. It seems peculiarly fitting that the origin of the
elements should be described by three men whose names
are the same as the beginning of the Greek alphabet.
Phi: The "Golden Ratio"
Pi crops up in many places in mathematics when we

least expect it. Another constant which has the same
pleasant property is known as Phi, or the Golden Ratio.
It can be defined in a great many ways, and perhaps the
easiest to remember is that it is the only number whose
reciprocal is exactly equal to one less than the number.
1/
Thus:
This leads to:
2 -
- 1
1= 0
= 1.61803398 ...
To illustrate some of the properties of "ubiquitousness," is the ratio between any line segment in a five
pointed star, and the next smaller segment in that figure.
Picture frames, playing cards, and other rectangles we
encounter seem to have a propensity for having their sides
in the ratio (i.e., they are Golden Rectangles).
One of the wildest attempts to show the ever-present
nature of was performed by F. A. Lonc, of N ew York. He
measured the heights of sixty-five women, and the heights
of their navels. (Why did he choose only women?) He reported that the average of a woman's height divided by
the height of her navel was equal to 1.618. It is said that
those women whose measurements were not close to this
ratio testified to hip injuries or other deformities.
* Physical Review, April 1948.
126
Martin Gardner reports that the same measurements

were performed by a group of men (each of whom measured
only his own wife) and got a value for this ratio about three
percent higher than cp. These men can justly claim that
they have "high-phi" wives!
Probability That Two Random Numbers Are
Relatively Prime, and Euler's Theorem
The fundamental theorem of arithmetic states that

each number has a unique decomposition into prime numbers. If two numbers have no common factors, they are
said to be relatively prime. Thus, 8 and 11 are relatively
prime, even though 8 is not prime itself. The probability
that a number is divisible by p is lip, because one out of
every p consecutive numbers is divisible by p: i.e., only
one in a string of five consecutive numbers is divisible by 5,
and only one in a string of seven is divisible by 7. Since
the probability that one number, chosen at random, is
divisible by p is lip, the probability that two numbers
chosen at random have the same common factor p is l/r,
and the probability that they do not have this common
factor is [1 - (1/p2)]. The probability that two numbers
chosen at random are relatively prime will be the infinite
product of terms of the form [1 - (1/p2)] where the p's
range over all the prime numbers. Thus, the probability
that two random numbers are relatively prime equals
;2)
~p~e8( 1 -
We will now show how to evaluate this product or,

more accurately, its reciprocal. We note first that
I
-- = 1 + x
I-x
+ x + x + ... xft + ...

2
where the absolute value of x is less than one.
127
We will take x to be equal to the 1/p 2 above. Then the

reciprocal of the probability is an infinite product of terms,
each of which is very similar to the expansion for 1/(1 - x).
1
II 1- (1/p2)
+ pl2 + Pl + Pl + ...)(1 + P2"2 +

P"24 + P"26 ... ) ... (1 + p;2 + p;4 + p;6 + ...)
= (1
In this product, every product of primes occurs in

unique fashion once and only once, and to the -2 power.
It therefore represcnts the sum of all integers (from 1 to
infinity) taken to the -2 power. It can be written as:
ex>
1 - p-2
= 1 + 2-2 + 3-2 + 4-2 + ... = L - 2

1
This last sum is the Zeta (r) function, and r(2) is known
to be 7r 2/6. The probability that two random numbers are
relatively prime is the reciprocal of this, namely 6/7r 2 If,
instead of expanding the infinite product II[1/(1 - p-2)]
we expand the product II[1/(1 - p-.)] then all our arguments could be modified slightly, and we would come out
with the result that:
all primes
1 - P-'
= r(s) =
ex>
n'
L-
This surprising result was originally discovered by Euler.

(That fellow sure got around, didn't he?)
15: A Slight Electrical Interlude

and Some Undigested Thoughts
There are a few problems whose solutions are consistent

with the aim of the rest of the book and which are based
on the very first principles of electric circuit theory. I have
been advised by some people that such problems do not
belong in this book. But an equally large and powerful
group feels that these problems must be included. To
satisfy both groups this chapter is included, but labeled so
that "electrophobes" can avoid it. For background, one
need know that a voltage E (volts) applied across a circuit
of resistance R (ohms) will cause a current I (amperes) to
flow, and that these quantities are related by Ohm's Law:
E = IR or volts = amperes X ohms.
Two-Dimensional Array of Resistors
Consider an infinite two-dimensional arrangement of

one-ohm resistors divided into small squares, as in figure 27.
What is the resistance between points A and B?
It is helpful to consider two simpler problems which
can be superimposed to give the desired solution. Consider
first, that one ampere of current flows in at point A, and
128
A Slight Electrical Interlude
129
I
--.&
-"
'9
:
A
TO INFINITY
ON ALL SIDES
,&
.....
.&
.&
Figure 27
that this current is taken out at infinity (or at a very large

circular ring, if you prefer). The current will surely divide
into four equal parts, one in each of the four equal resistors
leading from A . Now consider that the polarity is reversed,
and that the current enters at infinity and is extracted at
B. Again, the current must come in equal amounts through
the four resistors terminating at B. We can superimpose
these two solutions, applying a positive voltage E at A,
and a negative E at B. (The circle at infinity will correspond to zero voltage, and the currents flowing in and out
will cancel.) Of the current I which enters at A, 1/4 will
flow in the resistor between A and B. And of the current I
which is extracted at B, 1/4 again flows in that resistor.
A grand total of 1/2 flows in that resistance R when these
are superimposed, and causes a voltage E = IR/2. The
effective resistance is thus: E/I = R/2.
Q.E.D.
130
Ten Wires Under the River
Consider a ten-wire cable under a river. The wires are

not color-coded or marked in any way, and are available
for inspection only on either side of the river. It is desired
d
(a)
( b)
Figure 28
131
to tag and identify each of the wires. The only implements

allowed are a batch of clipleads (with which any wire
terminal may be connected to any other terminal, or group
of terminals) and an "on-off" continuity meter (with
which the presence or absence of a continuous circuit may
be determined.) How can all the wires be identified by one
man with only one round trip across the river?
There are at least two solutions to this problem. The
first one is valid only when the number of wires in the
cable is a triangular number (the sum of successive integers starting with one). To solve the problem according
to this first method, we consider the wires to be labeled
and connected as is shown in figure 28(a).
We connect wires in groups of 1, 2, 3, etc. On the
other side of the river, we can discover which wires are in
groups of 1, 2, 3, etc., but so far only wire 1 is positively
identified as being wire a. (We know, for example, that the
group 4, 5, 6 is the group d, e, I, but we can't separate it
any more than that.) Now we connect the wires, as is
shown in figure 28(b), and come back to the first side of
the river. Then we can again discover that wires are connected in groups of one, two, three, etc., and the fact that
each wire is uniquely determined as a member of a group
of m wires on one side, and of a group of n wires on the
other side, serves to identify each wire.
If the number of wires is not a triangular number, but
is an odd number, a different technique must be used. If
there were, for example, nine wires in the cable, we could
connect all the wires, save one, in pairs as is shown in
figure 29(a).
On the other side, all the wires save one are discovered
to be grouped in pairs. Then, since wire 1 is not connected
to anything else, it can be identified as a. The remaining
wires are connected in pairs as is shown in (b). Back on the
first side of the river, wire 1 (or a) is known to be connected
to b, and finding which wire is now shorted to number 1
identifies the wire which goes to b. The other wire of that
132

4
(0)
(b)
Figure 29
pair is known to be c, and the identification procedure

continues in this way until all wires are known. If the
number of wires is neither odd nor triangular, we can use
the odd-wire technique and never connect the extra even
wire on either side of the river. This extra wire which was
never connected to anything is thus identified.
Perimeter to Enclose Maximum Area
The geometrical figure which encloses the maximum

area for a given perimeter is well known to be the circle.
The following demonstration provides a simple, but intuitively satisfying, beginning of a proof. Consider a
smooth curve of arbitrary shape as is shown by the solid
curve in figure 30:
133
Figure 30
.. .,---...... At any indentation, such as the one between points A and

B, we can draw a straight line tangent to the curve at those
points, and reflect the curve in that line, getting the

dotted curve shown. The perimeter of the changed curve
is precisely the same as that of the original curve, but the
enclosed area has been increased. This reflection procedure
can be carried out so long as there is an indentation in the
curve, and eventually a smooth curve which has no indentations must result. This curve might be as shown in
figure 31:
Figure 31
Two points exactly half the perimeter apart are

marked. The line joining them can be called the pseudodiameter. In general, one of the two areas on opposite
sides of this line will be larger. We can replace the smaller
area with a mirror image of the larger area, and thus arrive
at an area larger than the original, but still of the same
134
perimeter. If the smooth curve is not perpendicular to the

pseudo-diameter, then there will be a dimple, or indentation, in the new curve. And then the process of increasing
the area, but not the perimeter, can be continued as before.
Only in the case where the curve is everywhere perpendicular to the pseudo-diameter can we no longer increase the
area without increasing the perimeter. But this latter case
defines a circle. So the circle must be the curve we sought.
Major shortcomings of this demonstration are that
it is not constructive, and it does not adequately deal with
angular figures, which may arise from reflection in the
pseudo-diameter. The beginning of the proof is presented,
however, in the hope that some reader may be able to continue along these lines, and arrive at a proof of this chestnut
in rigorous fashion.
16: Left-Overs-from Greeks to Computers
A Computer Correlation
The first digital computer was unveiled right after

World War II. In the succeeding years, there was a period
when new computers were appearing on the market faster
than they were being used to solve problems. During that
blissful era, a computer could be found for almost any interesting problem. In particular, one fellow fed vast
amounts of astronomical and geophysical data into a
machine, and asked the machine to correlate this data.
The machine did its bidding, and came up with many
interesting correlations. One of them was that every time
the fifth moon of Jupiter was in opposition to our moon, it
rained in Montreal, Canada. The fellow who was running
this program realized how very foolish it was to think that
the position of Jupiter's moon relative to our moon could
affect the weather in Montreal. It was more likely, he
thought, that the rain in Montreal, Canada, caused the
fifth moon of Jupiter and our moon to go into opposition.
Translation of Languages hy Computer
Computers are very much in the news today. Many

self-appointed prophets would have us believe that the
world will shortly become unrecognizable by our present
standards. There is little doubt that many new uses for
computers will be discovered, and that life will be changed.
135
136
LeftOvers-from Greeks to Computers
But the indiscriminate application of computers might

have drastic consequences-such as is indicated by the
following story:
An engineer had devised a program for a computer to
translate anything between any two languages on earth.
He reported this development at a technical meeting, and
to demonstrate the power of his program, asked the
audience to suggest a phrase for the machine to translate.
Someone suggested" Out of sight, out of mind." He entered
this into the computer, and then asked the audience for a
language into which he could translate the phrase. Another
person suggested Russian, and the computer was instructed to translate "Out of sight, out of mind" into
Russian. The machine whirred for a few moments and
announced that the required translation was
CJIEnOYf I1.lU10T.
Unfortunately, no one in the room understood Russian.

So no one knew whether it was a good translation or not.
Finally, someone had the brilliant idea that this Russian
could be fed into the machine with instructions to translate
it back to English. This was done, and the machine printed
that "Out of sight, out of mind," translated into Russian
and back into English, resulted in the phrase "Blind idiot!"
This story can also be told where the sentence suggested is "The spirit is willing, but the flesh is weak." This,
translated into Russian, and that Russian translated back
into English, becomes "The whiskey's O.K., but the meat
is lousy."
The difficulties that computers encounter in translating languages are apparent.
Fair Division of Cake Among N People
If there is a piece of cake to be shared by two people,

it is surely fair to let either person cut the cake and allow
the other person to choose his piece first. In this way,
137
neither person has any right to complain. Generalize this

fair procedure for N people, so that no one can complain,
regardless of collusion among any of the N people.
Anyone of the N people is asked to cut for himself a
piece of cake with which he would be satisfied. The remaining people are then asked if any of them think the
first man has too large a piece. If there are any who do,
one of them is asked to cut from this piece of cake a slice
such that he (the person cutting) would be satisfied with
the piece left. If no one feels that this piece of cake is too
large, then this second man may have that piece of cake.
And he cannot justly complain. If anyone, or any group,
does object, the procedure can be carried on until no one
does. The entire procedure is then repeated with (N - 1)
people, (N - 2) people, etc., until everyone has his cake.
The advantage of doing this problem with orange juice
instead of cake should be considered.
The Trolley Cars: An Apparent Violation
of Probability
A young man has two girl friends who live in diametrically opposite directions from his house, and whom
he normally sees about an equal number of times each
month. The trolley line which goes to both of their houses
pa.'lses in front of his house. Since he has an equal regard
for both of them, he decides to leave his house at completely random times, and to take the first trolley that
comes along. In this way, he thinks he will continue to see
each of them an approximately equal number of times,
since the trolleys go in each direction with the same regularity. He follows this regime for a few months and, much
to his consternation, he finds that the laws of probability
are apparently being violated. He has gone to see one girl
friend five times as often as he has gone to see the other!
How can this be?
138
In order to see how the laws of probability can apparently be violated, it is convenient to picture the track
along which the trolleys run, as is shown in figure 32a;
CAR BARN
I
I
I
I
Figure 32a
The trolleys run with perfect regularity in both directions, and because of this, there will be certain placeR
where two cars going in opposite directions always pass
each other. These are at points marked A and B. Assume
further, that the young man lives at the point marked x,
which is 1 minute from B, and 11 minutes from A. We
start our observation at the moment when the cars pass
each other, and if the man appears at x within the next
minute, he will meet a car going from B to A. If he appears
during the next 10 minutes after that, he will have missed
the car going from B to A, but will be in time for the car
going from A to B. One minute after the car from A passes
x, it will have reached B; the cycle will start over again.
If the man appears during this last minute, he will have
missed the car from A to B, but can wait for the next car
from B to A. Table 14 shows the situation during the
entire cycle;
Time
Situation
0-1 min.
1-11 min.
11-12 min.
next car will be from B

next car will be from A
next car will be from B
Table 14
Left-Overs-from Greeks to Computers
139
Thus, the next car will come from B only during two
minutes of each cycle_ But during ten minutes of the cycle,
the next car will be from A. If the man comes out at random times, he is five times more likely to go toward B than
he is to go toward A.
Napoleon's Problem
A curiosity which demands inclusion in a book such

as this is sometimes known as Napoleon's Problem (I've
never found out why). This problem asks for the construction with compass alone of two points which, with a
given pair of points, would be at the four corners of a
square. This problem is ridiculously easy with ruler and
compass, but can be tantalizingly difficult with compass
alone.
Figure 32b shows the two given points labeled A and
B. With radius AB, a circle is drawn about B, and an arc
around A. Point C, which is diametrically opposite A, is
found by the usual method of dividing a circle into six
Figure 32b
140
parts. * Then, with radius GE, the arc EF is described

about G, and likewise, with radius AD (the same as GE),
the arc DG is described. These two arcs intersect at point
P and the line BP is perpendicular to AG. Triangle AEG
is a right triangle because it is inscribed in a semicircle, and
the right angle is angle AEG. Thus, we have the equation
(AE)2 + (EC)2 = (AC)2, or in terms of r we can write:
r2
+ EG2 =
EG
(2r)2
= 4r2
= AB V3 = r V3
Two additional arcs as in figure 32b can then be drawn

because the length BP is the diagonal of the desired square.
The Mohr-Mascheroni Construction
In 1797 Mascheroni showed that all constructions

possible with ruler and compass were possible with compass alone. A Danish mathematician named Mohr had
done approximately the same thing in 1672. These constructions are now known as the Mohr-Mascheroni constructions. It is understood that straight-line figures cannot
be drawn, but the points defining these figures can be
constructed.
To prove that all constructions possible with ruler and
compass are possible with compass alone, it is sufficient to
show that a compass alone can be used to determine any
point which could have been determined with ruler and
compass. With both instruments we can determine the
intersections between:
* On a circle of radius T, exactly six consecutive and contiguous

chords of length r can be drawn. Each of these chords with two of
the radii of the circle will form an equilateral triangle. Each angle of
the equilateral triangle is known to be 60, and six 60 angles are
equal to 360, or a full circle. A circle can thus be easily divided into
six sections with a compass, by stepping off the radius six times
around the circumference.
LeftDvers-from Greeks to Computers
141
(a) two circles

(b) a circle and a straight line which does not pass
through the center of the circle

(c) a circle and a straight line which does pass through
the center of the circle
(d) two straight lines
In all of these cases, the straight line is defined by two
points on it.
-A
--~
I
\
, "'" -
/
./
Figure 33
Construction (a) above is trivial. The circles are

drawn and the points of intersection are evident. Construction (b) is shown in figure 33. The straight line is
defined by points A and B. The circle is centered at C and
is of radius r. The point C', which is the image of point C
in the line AB, is found by drawing arc CC' centered at A
and the other arc CC' centered at B. A circle of radius r
centered at C' will intersect C in the two required points,
E and F.
142
Construction (c) is more complex; it is outlined in

figure 34:
Figure 34
The line AB is indicated by the dotted line in figure
34, and the circle of radius r is centered at A. With arbitrary radius BC, arc CD is drawn centered at B. The
problem is to bisect arc CD. Arc AE is drawn centered at
C with radius r, and arc AF is centered at D with the same
radius. Length CD is measured on arcs AE and AF,
defining points G and H. Line AG is parallel to CD as is
AH, because both AGCD and AlIDC are parallelograms.
Arcs DP, centered at G, and CP, centered at H, are drawn,
intersecting at P. Length GD is the hypotenuse of a right
triangle whose base is I CD and whose altitude is
yr2-
(CDj2)2
The length GD is thus:
yr
2 -
t (CD) 2 + t (CDF = yr2
+ 2(CD)2
The length AP is then:
y(GD)2 - (AU)2 =
yr + '2(CD)2 yr2+ (CD)2

2
(CD)2
The point x which is the midpoint of arc CD is a distance
143
vr2
+ (CD)2 = AP from point G and thus can be constructed.

Construction (d) is by far the most difficult of the four.
Figure 35 shows the four points A, B, C, and D, and the
lines connecting them are shown dotted:
__
_--
-- ---- --_x_----- --_--
--
c
A
0--
c'
----~~............
~E
I
I
I
~
B
....
Figure 35
Points C' and D' are the images of C and D in line AB.
Point E is now constructed to be the fourth point in parallelogram CC'D'E. Since CC'is perpendicular to AB, D'E
is also. And since DD' is also perpendicular to AB, ED'D
must be a straight line. In triangle DCE, DX/DC =
DD'/DE. The length DX is the only unknown in this
equation. It is called the fourth proportional. It can be
constructed as shown in figure 36.
The circles of radius DD' and DE (both known) are
drawn with the same center. At an arbitrary point P on
the smaller circle, the arc RS is drawn with radius DC. The
arc TU is drawn about S with the same radius DC. The
length RT is the fourth proportional, i.e., is the length
DX which we sought to construct. This is so because
LROT = LPOS (LPOR = LSOT and LROS is
added to both of them). Chords PS and RT subtend
the same size angle at the center and they are, therefore,
proportional to the radii of their circles. Having length DX,
144
LeftOvers--from Greeks to Computers
Figure 36
we go back to figure 35 and describe arcs DX and D'X,

thus defining point x. The construction is now complete.
And since the ingredients of a construction with ruler
and compass have now been done with compass alone, the
proof of the Mascheroni Construction is complete.
"Mohr" of these Mascheroni constructions are described in a booklet by A. N. Kostovskii. *
Random Chord of a Circle
What is the probability that a random chord of a

circle is longer than one side of the inscribed equilateral
triangle?
A random chord will be determined by two random
points on the circumference of the circle. An inscribed
equilateral triangle can be rotated inside the circle so that
one of its apices coincides with one of the points. Then,
if the second point falls between the other two apices, the
random chord will be longer than a side of the triangle.
* Kostovskii, A. N., Geometrical Construction Using Compass

Only. New York: Blaisdell Scientific Publications, 1961, vol. 4 of
Popular Lectures in Mathematics Series.
145
The available circumference to get the longer chord is !

of the full circle, so the probability must be !.
But there is another way to solve this problem. We
can say that a random chord is determined by a point on
a randomly oriented radius vector of the circle. Wherever
this point falls, we draw a line perpendicular to the radius
vector, and all possible chords can be represented this way.
Only if the randomly chosen point is within! of the radius
of the circle will the random chord be longer than the side
of the triangle. The random point can fall anywhere along
the radius, and since only the centermost half yields a
favorable result, the probability that the chord is longer
must be!.
But hold! What have we here? (Enter guards with
clanking armor to remove author.) Two different answers
for the same mathematical problem? Let's try a third way.
And if the third way comes out the same as either of the
other two, then we will at least have the weight of probability on our side. A random point in the interior of the
circle will determine a random chord because we can draw
a line from the point to the center, and then a perpendicular to that line determines the chord. But the point
must be within a circle of radius r/2 in order that the
chord be longer than the side of the triangle. The probability of getting a randomly chosen point within the
center circle of ! the full circle radius is obviously proportional to the area, and hence the required probability is t.
This demonstration shows rather clearly that words
like "a randomly chosen chord" must be carefully considered and defined.
Description of Greek Names for Numbers
The Greeks had many names for numbers. Numbers

were not simply odd or even, or large or small, but they
were deficient, perfect, sociable, friendly, abundant,
square, etc.
146
LeftDvers-from Greeks to Computers
Perfect numbers are those for which the sum of all the
divisors of the number (including one, but excluding the
number itself) is equal to that number. Thus, 6 is perfect
because its factors, 1,2, and 3, have the sum 6. The factors
of 28, 1, 2,4, 7, and 14, have the sum 28, and so 28 is also
perfect. Perfect numbers are very well understood, and
there is a formula 2n - 1 (2n - 1) where (2n - 1) is prime for
all the perfect numbers known so far. (A tantalizing question is, "Are there any odd perfect numbers?")
Deficient numbers are those in which the sum of the
divisors is less than the number itself, and abundant numbers are those in which the sum is greater. Examples of
deficient and abundant numbers are not difficult to find.
Friendly (or amicable) pairs of numbers are those in
which the sum of the divisors of each number is equal to
the other number, as for example, 284 and 220. The divisors of 220 are 1, 2, 4, 5, 10, 11,20, 22, 44, 55, and 110which total 284; the factors of 284 add up to 220.
In sociable chains the sum of the divisors of each
number is equal to the next number, and the sum of the
divisors of the last number is equal to the first number.
(It requires what the Germans call" sitzfleisch" to discover
such chains, but number theorists are apparently willing
to devote time to such problems.) An example of a sociable
chain is 14,288, 15,472, 14,536, 14,264, 12,496. (I can't bring
myself to verify this, or to ask any of my friends to do so.)
Some Geometricians Are Masochists
The usual picture of a mathematician is that of a shy
individual, often with glasses and a receding forehead (or
a large scalp.) Geometricians, in particular, have peculiar
ways of torturing themselves. It is known, for example,
that Euclid made many geometrical constructions using
only a straightedge and a compass. But some geometricians
go still further-they try to use less than the standard
straightedge and compass.
147
For example, Mohr and Mascheroni showed how

many constructions could be accomplished with a compass
alone. Steiner* showed that Euclid's constructions could
be accomplished with a single circle and a straightedge.
Desargues' Theorem and the Wilkes-Gordon Constructions
show how the straightedge alone can be used to accomplish
certain difficult constructions.
A problem was recently reported in which one is required to find the perpendicular bisector of the line joining
2 points, using only a tri-square. A tri-square is an implement which can draw straight lines and erect a perpendicular at any given point on a line, but it cannot be used to
drop a perpendicular to a line.
There have been so many examples in which seemingly idle amusements turned out subsequently to have
great and useful applications, that it would be foolhardy
to criticize geometricians for limiting the tools available to
them. (But only some mathematicians consider practicality to be a desirable feature. For example, it is said
that Kummer, the mathematician, was once asked which
of all his discoveries he appreciated most. He replied that
he appreciated ideal numbers most, because they had not
yet been soiled with practicality.)
Ancient, Admirable Calculations
Figure 37 shows a pair of intersecting cylinders, both

of the same radius r. What is the volume common to both
cylinders?
This problem can easily be solved by integral calculus,
but the answer was known to Archimedes, who antedated
the discovery of calculus by many years. What methods
did he use?
We don't know the method used by Archimedes, but
it may have been similar to the following:
* Courant and Robbins (loe. cit.), p. 197.
148
Consider a sphere of radius r to be at the center of the

volume common to both cylinders. If we make a series of
slices parallel to the plane of both cylinders, we will find
that each thin section of the volume common to both
cylinders is a square; and that a portion of the sphere is
present in each one of these slices as a circle inscribed in
that square. Each slice, then, consists of a square from the
- --
~
I
I
I
--
I
I
RADIUS
!.
()
J
Figure 37
common volume and an inscribed circle from the sphere.

The area associated with the sphere in each slice in 7ra 2,
and the area associated with the common volume is (2a)2,
where a represents the radius of the inscribed circle. The
ratio of area in each slice belonging to the common volume
149
to that belonging to the sphere is 4a 2j1l"a 2 The total

volume, which is the sum of all these slices, will be in the
same ratio of 4j1l". So the common volume must be
A similar problem asks for the volume of the solid

shown in figure 38. The solid has a circular base, and a
"straight-line top." The sides can then be considered to be
straws starting at the top line and going to the circumference of the circle at the bottom but always perpendicular
to the line at the top. What is the volume of this figure?
Figure 38
Again, calculus can be used to get the answer, but we

ask here for the answer without the aid of calculus. Figure
39 shows how to calculate this volume. Again, we slice the
figure in a plane perpendicular to the line at the top. Each
thin slice will have the shape of a triangle. If we compare
the given figure with a right circular cylinder of the same
base and same height, each slice of figure 39 is exactly
one-half of the comparable slice of the right circular
cylinder. The entire volume of figure 38 is exactly half of
the volume of the comparable right circular cylinder.
The volume is thus equal to !1/"T 2L.
150
Left-Overs-from Greeks to Computers
Figure 39
V"'"
\ ......
Fermat's Challenge
The mathematicians in a bygone era had one very

interesting characteristic which seems to have disappeared.
They used to challenge each other or all mathematicians of
a particular nationality, to do a particular problem. For
example, in 1657 Fermat posed the problem of finding a
cube which, when increased by each of its divisors, is equal
to a square. One answer is 73 + 1 + 7 + 7 2 = 400 = 20 2
On another occasion, he challenged the English mathematicians to find all integral solutions of the equation:
x 2 - ay2 = 1.* This challenge, incidentally, was correctly
answered by Lagrange in 1767. It seems a pity that this
method wasn't developed to provide a means for settling
problems between nations.
Dance of the Bugs
We have calculated the probability that two numbers

chosen at random are relatively prime. Nowhere in this
proof was it necessary for us to define exactly what we
* Beiler, Albert H., Recreations in the Theory of Numbers. New

York: Dover Publications, pp. 248 ff.
151
mean by a random number. But in spite of this, the calculation of the probability was performed. (I even think it
would satisfy the mathematical purist.) The present
problem asks for a determination of the length of a very
peculiar curve. We can solve this problem without determining the shape of the curve.
Consider a square 10 inches on a side, as is shown in
figure 40. A bug waits at each corner of the square and at
a given signal (these are trained bugs) each one begins to
walk toward the one in front of it, as is shown by the
arrows. As each bug moves, the direction of the one behind
him changes, because the bugs always move toward the
one in front. They move at a velocity of one inch per
minute. How long does it take before the bugs collide with
each other in the center of the square, and how far has
each one walked?
In this problem also, we have an apparent dearth of
information. To solve it, we need recognize only that the
travels of bug B are always at right angles to the path
from bug A to bug B, and so cannot affect the distance
from A to B. Then it is only A's walking that has any
effect on the distance from A to B, and only B's travels
affect the distance from B to C. Since the bugs start
10 inches apart, they can travel 10 inches before a central
collision. Each bug must travel a distance of 10 inches,
10"
10"
Figure 40
152
LeftOvers--from Greeks to Computers
and, at a velocity of one inch per minute, this will require

10 minutes. But the shape of the curve that the bugs
travel is still not determined.
The Swiss Mountain Moving Company, Ltd.
(I leave it to the reader to decide whether the third
word in the title is used in an adverbial or nominal sense.)
A group of movers were accustomed to using rollers
on which to transport heavy objects, but these rollers kept
rolling downhill. The men tried using square rollers, but
Figure 41
found the up-and-down motion of the object being moved

extremely objectionable. They consulted an oracle to find
out if it was possible to design a roller which would not roll
downhill of its own accord, but on which a heavy object
could be moved smoothly. The oracle (who was a toothless
old hag, as are all respectable oracles) suggested that there
were many curves which had a "constant diameter"
(the diameter of any closed curve is defined as the longest
straight line that can be drawn inside the figure) but which
were not circular. The fact that they weren't circular would
prevent them from rolling downhill. And the fact that they
153
were of constant diameter would allow objects to roll

smoothly.
What could have been the cross-sectional shape of
these rollers?* The drawing in figure 41 shows one such
curve. At each of the comers of an equilateral triangle, a
small circle of radius ro is drawn. And from each of these
points, an arc of radius rl is drawn tangent to the other
two circles. A roller with this cross section will not roll
down a gentle hill. But an object placed on top of such a
roller will remain at a constant distance above the ground.
* A West Coast electronics firm uses such rollers as an advertising

gimmick.
Index
"Advantages of Compound Interest" The," 41-42

"aG - a Is Divisible by 30,"
57-58
ages, problems about, 1-7

Alpher, R. A., Dr., 124-125
American Mathematical
Monthly, 17,47n, 118n
Amthor, A., 119
anchor ring, volume of, 25-26
"Ancient, Admirable Calculations," 147-149
"Another Coin Problem," 47-48
"Another Urn Problem," 64
Archimpci('s, 121, 147
"Archimedes' Cattle Problem,"
119
area
enclosed in a perimeter,
132--134
of squares, 33-35
of triangle, 40-41
area problems, 33--35, 79-80,
102--103, 132--134
arithmetic
Fundamental Theorem of,
60, 126
see restored arithmetic
"Basketball Scores," 12--13

Bell, A. H., 119
Bell, E. T., 122
Bernoulli numbers, 54
"Bertrand's Box Paradox," 107108
Bethe, Hans, Dr., 124-125

Bible, 121
binary number, 74
binary representation, 67-68
binary sequence, 43
birthday problem, 110
Buffon, Count, 112, 123
cake-cutting problem, 136-137
"Calculation of Pi to 100,000
Decimals" (Shanks,
David, and Wrench),
122n
Candide (Voltaire), cited 123n
"Careless Indian Who Drops
His Paddle, The," 99
"Casting Out Nines," 90--91
catalyst, 119-120
Chartres, R., 123
check-cashing problem, 13-14
"Cigars on Table," 72-73
coin problems, 43-48, 107-108,
110--113
Bachet's weight problem, 43

Ball, W. W. R., 9-10
compound interest problem,

41-42
155
156
Index
computers, 135-136
constant diameter, 153
construction
problems, 37-39,80-81, 139140, 147
see Mohr-Mascheroni construction; Wilkes-Gordon construction
"Construction of a Cube," 80-81
Count Buffon's Needle Problem, 123-124
counters, games played with,
74-79
Courant and Robbins, cited 37n,
147n
cube
construction of,80
problems, 10-12, 150
when equal to a square, 150
"Cup of Coffee and Cup of
Tea," 73-74
cylinders, intersecting, 147-149
"Dance of the Bugs," 150-152

decimal, repeating, 67
decimal expansion, 31-32
decimal number, 74
"Definition of a Catalyst," 119120
denominator, greatest common,
69
Desargues' Theorem, 37-39,147
"Design Tracing," 20-22
"Determination of Odd Coin
Among Twelve Coins,"
43-47
"Die Throwing," 106-107
digital,90-91
Diophantine-Type problems, 917
equations, 8-10, 13
disc-stacking problem, 30-31
divisibility, 57-59, 126

partitioning problems, 6170
by various numbers, 92-98
"Divisibility of a 13 - a," 58-59
division, restored, 10
"Division of Balsam Among
Three People," 63-64
division problems, 61-70, 136137
"Early Commuter, The," 99-100
"Eight 8's," 82-84
electrical problems, 128-129,
130-132
Euclid, 146-147
Euclidean Algorithm, 61~2, 6465,69-70
Euler, Leonhard, 18-20, 49,
117-118, 123
formula of, 23-24
<p function of, 51
theorem of, 126-127
"Expression for Which 2,592
is a Solution," 15-16
factorial function, 82-84
"Fair Division of Cake Among
N People," 136-137
Fermat, Pierre, 49-50, 150
"Fermat's Challenge," 150
Fermat's Theorem, 49-60
Fermi, E., 55, 56
"Finding Any Rational Fraction of a Volume V,"
6~8
games, symmetry, 71-73, 74-79

gamma function, 82, 83
Gamow, George, Dr., 120-121,
124-125
Gardner, Martin, cited 22n, 79,
126
Index
157
Konigsberg Bridges, 18-20

Kostovskii, A. N., 144
Kummer, 54, 147
magic tricks, 88-90

"Maid at the Stream, The,"
61-63
map, surface of, 23-24
Mascheroni, 140, 144, 147
"Master Painter VB. the Apprentice, The," 102-103
"Mathematical Education in
the
USSR"
(Gnedenko), 47n
mathematical expectation, 110113
Mathematical Recreations and
Essays (Ball), IOn
Milburn, R., Professor, 80
"Missionary and the 159-Link
Chain, The," 9
mnemonics, 120-121, 122-123
modulus, 50-51
Mohr, 140, 144, 147
Mohr-Mascheroni Construction,
The, 140-144
moving problem, 152-153
MUltiplication Theorem, The,
84-91
"Mutilated Chess Board," 7980
Landon, V. D., cited 102n

Lagrange, 150
Lazzerini, 124
"Liar or Truth-Teller-at Ntuple Intersection," 116
"Liars and Truth-Tellers," 115
Life magazine, 55
Lindemann, 122
liquid measurement problems,
61-64, 6IHi8
logic problem, 32-33
Lone, F. A., 125
loop around a torus problem,
27-29
Loyd, Sam, 71
"Napoleon's Problem," 139-140

needle problem, 123-124
"Neighborhood Census, The,"
1-3
"Nim," 74-77
"Nim.," 77
numbers
abundant, 145
Bernoulli, 54
binary, 74
composite, 55, 118
cyclic permutation of, 3132
decimal,74
deficient, 145, 146
Gauss, 118
"Geometric Proof That
~ m3
[N~ mJ2"
,
33-35
Geometrical Construction Using

Compass Only (KoBtovskij), 144n
Gnedenko, B. V., 47
"Goldbach's Conjecture," 117
Golden Ratio, 125
Golden Rectangles, 125
Golomb, S., 33
Gordon, G., Dr., 37, 38n, 80
Green, M. W., 33
"Grundy's Game," 78-79
harmonic series, 31
"Hempel's Paradox," 32-33
"Incorrectly Cashed Check,
The," 13-14
"Integral Pythagorean Triplets," 16-17
158
Index
numbers (cont.)
even, 10, 12,22-23,75
friendly, 145, 146
Greek names for, 145-146
ideal,147
Ludolphian, 122
manipulation of, 82-91
odd,22-23,35,54,75
perfect, 145-146
prime, 35, 49-60, 123-124,
126-127
relatively prime, 13, 123,
126--127
sociable, 145, 146
square, 145
square of, 51
squares of, problem, 17
squaring of, 85-88
transcendental, 122
triangular, 119, 131
see also binary
number theory problems, 31-32,
54-59, 150
O'Beirne, T. H., 22
"Officer and His Column of
Men, The," 100-102
ohms, 128
Ohm's Law, 128
"On Moving Day," 14-15
Pappus' Theorem, 25
parallelepiped, volume of, 10-12
"Parity of Number of People
Shaking Hands at a
Conference," 22-23
"Peculiar Infinite Sequence, A,"
35-37
"Peculiarity of 142, 857," 31-32
"Perimeter to Enclose Maximum Area," 132-134

Phi, 125-126
Physical Review, 124-125
Pi,121-124
"Pile of Bricks," 65
Platonic solids, 24
Poisson distribution, 122-123
"Poker Chips," 65-6
polyhedra, 24
"Priest and the Banker, The,"
3-4
prime numbers, 35, 49-60, 117,
123-124, 126--127
probability
distributions, 122-123
problems, 86--87, 106--113,
123-124, 137-139, 144145
that two random numbers
are relatively prime,
126--127
"Probability of Identical Pairs
Among Cards," 108109
"Probability of Same Birthday
Among N People," 110
Proof of Fermat's Theorem,
59-60
Proof of the Pythagorean Theorem, 40-41
Ptolemy, 121
Pythagorean Theorem, 40-41
Pythagorean triangles, 16--17
"Railroad Trestle Problem," 16

"Random Chord of a Circle,"
144-145
Recreations in the Theory of
Numbers (Beiler), 150n
"Rectangular Block Cut into
Unit Cubes," 10-12
Index
rectangular parallelepiped, volume of, 10-12

relatively prime numbers, 13,
123, 126-127
repeating decimal, 67
"Restored Arithmetic Problems," 9--10
Rothlein, B., cited 100
"Sailors and Coconuts," 55-57

"St. Petersburg Paradox," 110113
"Sam Loyd's DaisY Puzzle,"
71-72
Schnirelmann,117
Scientific American, 22, 120-121
"'Seemingly Impossible' Problem, A," 118
sequence
infinite, 35-37
problem, 35-37
of remainders, 32
see also binary sequence
Shanks, 122n
Sleeping Beauty, 114
"Some Geometricians Are Masochists," 146-147
sphere
surface of, 23, 24
volume of, 105
square
construction, 139--140
unit: area of, 33-35
square root, restored extraction,
10
"Stacking of Discs," 30--31
Steiner, cited 37n, 147
"Story of Seven Bridges," 18-20
surface, 23-24
see also Euler's Formula;
arC':t
surface arC'a prohlem, 102-103
159
"Swiss Mountain Moving Company, Ltd., The," 152153

sYmmetry problems, 71--81
"Ten Wires Under the River,"
130-132
"Three Martians, The," 4-7
"Three Men and Twelve Lumps
of Sugar," 54-55
time problem, 99-100
"Topological Behavior of a
Loop Around a Torus,"
27-29
torus, surface of, 24
triad,3-7
triangles, 17,40--41, 140n
"Trolley Cars: An Apparent
Violation of Probability," 137-139
"21 Counters, The," 78
"Two-Dimensional Array of
Resistors," 128-129
"Two Ferry Boats, The," 103105
Van Ceulen, Ludolph, 121-122
velocity problems, 16, 99
Vieta, F., 121
Vinogradoff, 117
Voltaire, cited 123n
volume
common to intersecting
cylinders, 147-149
problems, 10-12, 25-26, 6164, 73-74, 104-105, 147149
of a rectangular parallelepiped, 10-11
of tube, 26
"Volume of an Anchor Ring,"
25-26
160
Index
"Volume of a Modified Bowling Ball," 104

Waite, John, 118
weight problem, 43-48
What Is Mathematics? (Courant and Robbins), 37n
"Where on Earth Was He?"
116-117
"Which
Raise Nets More

Money?" 29-30
Wilkes, R., 37, 38n
Wilkes-Gordon Construction,"
37-39,147
Wilson's Theorem, 50
Zeta function, 127

Mathematical Entertainments PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Mathematical Entertainments PDF

Uploaded by

Copyright:

Available Formats

M. H.

Thomas Y. Crowell Company, New York

All rights reserved. No part of this book may

Designed by Laurel Wagner

Library of Congress Catalog Card No. 65-15225

* Research Engineer, formerly at R.C.A. Labs., Princeton, N. J.

sense or another, are represented. Most of them can be

The Neighborhood Census, The Priest and the

liarity of 142,857, Hempel's Paradox, Geo-

Infinite Sequence, Wilkes-Gordon Construction, Proof of the Pythagorean Theorem, The

5: Coin Weighing Problems

Determination of Odd Coin, Weak and Strong

6: Fermat's Theorem; Sailors and Coconuts

Three Men and Twelve Lumps of Sugar,

The Maid at the Stream, Division of Balsam

Sam Loyd's Daisy Puzzle, Cigars on Table,

9: Parlor Tricks and Number Manipulation

Eight 8's, The Multiplication Theorem, Some

The Careless Indian Who Drops His Paddle,

14: Humor in Mathematics

Liars and Truth-Tellers, Liar or Truth-Teller

Letter in the Physical Review, Phi: The

Two-Dimensional Array of Resistors, Ten

A Computer Correlation, Translation of Languages by Computer, Fair Division of Cake

One of the few amusing things to come out of World War

What are those ages?

determine the three numbers. In this case, the census

The town priest saw his friend the village banker

The solution to this problem is very similar to that

We can see that only 5, 10, 49 and 7, 7, 50 have the

The last twist of this popular type of problem is, in

Wanting to know the ages of three Martians, the

But, after our passions have subsided, we decide to

the triad in question must have been 2, 10, 16. In the

Diophantine equations are those equations in which only

There is another solution to this equation in which x and

which leads to:

These are the only solutions (excluding the trivial x = 0,

The Missionary and the 159Link Chain

Restored Arithmetic Problems

The previous problem hardly requires Diophantine

gives an example* of a restored division problem with no

The next problem is: A rectangular parallelepiped,

* Ball, W. W. R., Mathematical Recreations and Essays, New

from each edge, or two units less in each dimension. We

= 2[(a - 2)(b - 2)(c - 2)]

An 8 X 10 X 12 original block will have an inner core

Dividing by abc, and inverting, we obtain:

may amuse himself by finding them if he wishes to. The

A basketball team thought one point for a free throw

playing around with this problem and several sets of

Our next problem is about the man who received a

must be approximately twice as large as r, so let us

peered in a window. Seeing the house completely empty,

This friend of mine from graduate school was nearly

* This is a true story. The phone number in question was the

Larry's method for discovering that 2,592 is a unique

Two boys were crossing a railroad trestle when they

The boys therefore ran at a velocity of 10 mph.

The discovery of integral Pythagorean triplets can be

value of c is given, there are infinitely many solutions for

All primitive Pythagorean triplets can be formed this way.

3: Konigsberg Bridges, Design Tracing,